LIVEPublicAPI accessHosted by@OpticFeeds

Crypto Blogs Tier 2 | Tracking

Stream tracking multiple Crypto blogs, including Medium.

Paid stream. Faster version can be found here:
https://synoptic.com/p/streams/01K0VJQT2F1PSGR2GD7Y66WDGV

cryptoblogs-2In packageDelayed 10 s for guests12

Connect your agent

Follow https://synoptic.com/p/streams/medium-mid/SKILL.md to connect Crypto Blogs Tier 2 | Tracking from Synoptic

Data stream

This is a delayed preview. Log in to follow the stream and see new posts in real time.

medium.com: Japan Wants To Put Its Government Bonds Onchain. Here’s Why That Matters<br /> Japan Wants To Put Its Government Bonds Onchain. Here’s Why That Matters Japan’s Ministry of Finance has started formal work on bringing Japanese government bonds, one of the largest pools of debt in the world, onto a blockchain. The ministry held its first study group on the idea on October 8, and it is now gathering industry feedback on how a tokenized JGB would actually work. The market noticed. A new report from the digital assets research desk of one of the world’s biggest banks says Japan’s plan could be the push that carries the tokenized real-world asset market to a forecast of $2 trillion by the end of 2028, up from roughly $45 billion today. Japan Wants To Put Its Government Bonds Onchain. Here’s Why That Matters What Japan Is Actually Considering The study group is weighing three paths. The first is to tokenize existing government bonds: the bonds stay exactly as they are in legal terms, and a token is issued that represents a claim on them. The second is to issue brand new bonds directly onto a distributed ledger, so the token is the bond rather than a wrapper around it. The third is a mix of the two. The difference matters. Wrapping an existing bond is the simpler option because the legal structure, the cash flows and the holder’s rights all stay the same. Issuing natively on a ledger is harder, but it removes whole layers of reconciliation, because the record of who owns the bond and the bond itself live in the same place. Over time that is where the real savings in issuance and settlement costs sit. Japan has not committed to any of the three. There is no chosen blockchain, no timeline and no decision on whether the first issuance would be a small pilot or something bigger. But a finance ministry study group is how these programs start, and Japan’s is now running. Why Long-Dated Debt Is The Missing Piece Almost all government debt onchain today is short-term. Tokenized US Treasury bills total around $13 billion, and they behave more like cash management tools than bonds. Tokenized government debt outside the US adds up to less than $1 billion. The long end of the curve, the ten, twenty and thirty year bonds that pension funds and insurers actually hold, is essentially untouched. That gap is why the $2 trillion forecast has looked ambitious. The projection leans mostly on tokenized money market funds and tokenized equities, each growing to about $750 billion. Longer-dated government bonds would give the market a third engine, and a big one: Japan’s government bond market is the third largest in the world. The appeal of a tokenized long bond is practical. Tokens can be issued and settled more cheaply, bought in smaller denominations, traded around the clock, and moved as collateral without waiting for legacy settlement cycles. For a market where trillions sit parked as collateral, that last point alone justifies the study group. Why Other Governments Are Watching Japan is not the first country down this road. The UK has a Digital Gilt Instrument pilot expected to go live in early 2027, and several smaller programs are running elsewhere. But pilots in small markets only prove the technology. A program in a market the size of Japan’s would prove the economics. That is the real signal in the research note. If tokenizing JGBs shows measurable savings on issuance and settlement, and makes bonds easier to move as collateral, the US and other major bond markets come under pressure to follow. Tokenization of sovereign debt stops being an experiment and starts being a competition. What Could Slow It Down The honest answer is almost everything, at this stage. There is no ledger chosen, no legal framework settled, and no decision on whether a tokenized JGB is a security on a blockchain or a blockchain record that is the security. Those are different things with different consequences for ownership, custody and settlement finality. Market plumbing pushes back too. Bond markets run on set hours and established settlement cycles, and a token that trades 24/7 has to reconcile with both. Liquidity is another question: a tokenized issue that only a handful of institutions can hold and trade is a pilot with extra steps, not a market. And the first version is likely to be the wrapper, not native issuance. Wrappers cut some costs but keep most of the old machinery underneath, which means the biggest benefits wait for the harder option. What This Means For Tokenization The pattern of the last two years has been tokenization climbing the quality ladder: first treasury bills, then money market funds, then private credit, and now the core sovereign bond markets. Each rung brings in more conservative holders, and each one demands cleaner infrastructure. For issuers the requirements do not change with scale: an accurate record of ownership, transparent backing, and compliance rules that travel with the asset from issuance onward. That lifecycle is what TokenFi’s real-world asset platform is built to manage, from issuance through ongoing compliance. TokenFi is not involved in Japan’s program. The thing to watch is what the study group recommends next year. A decision to tokenize existing JGBs is meaningful. A decision to issue natively onchain would be the largest commitment any major government has made to this market. Connect With TokenFi TokenFi is building the infrastructure for the tokenized future, making it easy for individuals, creators, and institutions to launch tokens, tokenize real-world assets, and participate in the next evolution of on-chain finance. Stay connected through the official channels below to explore TokenFi’s ecosystem, access documentation, and follow the latest product launches and updates. TokenFi website: https://www.tokenfi.com/ TokenFi on X: https://x.com/tokenfi TokenFi Whitepaper: https://docs.tokenfi.com/ TokenFi RWA: https://www.tokenfi.com/rwa TokenFi Launchpad: https://lp.tokenfi.com/en-US TokenFi Token Launcher: https://www.tokenfi.com/token-launcher TokenFi Nimbus: https://www.tokenfi.com/nimbus TokenFi Shield: https://www.tokenfi.com/shield TokenFi AIGen Studio: https://www.tokenfi.com/aigenstudio Link

news.curve.finance: Curve Selected Yields & Key Metrics: Week 47, 2026<br /> Curve this week in 60 seconds The PYUSD/crvUSD pool paid 25.8% APR over the last 7 days, mostly in crvUSD rewards that YieldBasis pays out on Merkl . The eUSD/crvUSD pool paid 23.4% , mostly in CRV. The BOLD/LUSD and BOLD/USDC pools pay 53.0% and 39.5% APR right now, after a one-off jump in BOLD rewards from Liquity that runs until November 26. scrvUSD now pays 7.9% APY, up from 6.8%, and deposits grew from 17.6M to 19.0M crvUSD. Two new Llamalend v2 markets take wOUSD and wstGBP as collateral, and you can supply crvUSD to them. A DAO vote would move the YieldBasis WETH, WBTC, cbBTC and tBTC pools to Pyth price data , so the pools keep more of the trading profit. Selected yields Markets shown have at least $100,000 in pool liquidity or supplied assets. Pool tables show two rates for a new $100,000 deposit: what LPs earned over the last 7 days (7d APR) and what the pool pays today (Now). Both count trading fees and unboosted rewards. Lending and looping rates are today's rates. Rates change often, and this is a selection, not a full ranking. CRV rewards can be boosted via Convex , StakeDAO or Yearn . Selected USD yields Chain Market 7d APR Now BOLD LUSD 11.8% 53.0% BOLD USDC 9.2% 39.5% PYUSD crvUSD 25.8% 18.0% eUSD (Reserve) crvUSD 23.4% 12.5% sUSG reUSD (Re Protocol) 9.7% 10.1% frxUSD USD3 (3Jane) 9.7% 9.4% scrvUSD sDOLA 8.0% 7.8% reUSD (Re Protocol) sUSDe 7.6% 7.6% apyUSD apxUSD 7.4% 7.4% scrvUSD sUSDe 7.4% 7.4% StableSwap pools with more than 85% of value or normalized reserves in one coin are omitted. FXSwap pools, such as zCHF/crvUSD, are assessed separately from this stablecoin-parity cutoff. Why BOLD/USDC pays more this week Most of the 39.5% APR the BOLD/USDC pool pays right now comes from BOLD rewards, and it is a one-off. Over the last 7 days the pool paid 9.2%. Liquity sends 25% of the interest its borrowers pay, including the fee charged when a loan is opened, to its governance contract. LQTY voters then split that money each week between incentive programs. During the sBOLD vault exploit on November 18, the attackers opened and closed loans worth millions of BOLD, and each loan paid that opening fee. Governance took in 104,385 BOLD during the attack, against 258 to 1,037 BOLD on a typical day. The epoch that closed on November 19 sent 62,452 BOLD to the pool's gauge, while a usual week brings 5,100 to 6,400. These rewards stream until November 26. A normal epoch would add about 3%. Llamalend v2 looping yields Chain Collateral 20x Looped APR Liquidity sDOLA 48.3% 2.22M sfrxUSD 9.3% 1.74M Example: $10,000 looped 20x to $200,000 of collateral with $190,000 crvUSD borrowed. Looped APR = 20 × 7-day collateral APR − 19 × borrow APR after that borrow. Selected Llamalend supply yields Chain Lent asset Collateral APR crvUSD reUSD (Resupply) sfrxUSD 14.1% crvUSD svZCHF 5.1% crvUSD WETH 5.1% crvUSD WBTC 5.0% crvUSD wstETH 4.9% crvUSD sfrxUSD 4.2% crvUSD sUSDe 4.1% crvUSD sDOLA 3.9% Lending APR may include borrower interest and CRV rewards. Supplier exposure includes the borrower collateral shown, and withdrawal liquidity varies with utilization. Selected BTC and ETH yields Chain Market Asset 7d APR Now WETH pufETH ETH 6.4% 6.7% pufETH wstETH ETH 4.0% 5.1% ETH+ WETH ETH 3.3% 3.2% WETH weETH ETH 2.9% 2.9% osETH rETH ETH 2.3% 2.3% tBTC cbBTC BTC 4.3% 3.5% WBTC cbBTC hemiBTC BTC 3.0% 2.6% Other selected yields Chain Market Asset 7d APR Now WETH CVX CRYPTO 11.3% 10.6% frxETH CVX CRYPTO 10.1% 9.3% crvUSD WETH CRV TRICRYPTO 5.4% 5.1% crvUSD tBTC wstETH TRICRYPTO 4.7% 4.6% USDT WBTC WETH TRICRYPTO 4.6% 4.5% Protocol activity and stability crvUSD & scrvUSD Metric Value Vs W46 Mint-market debt $40.2M -20.0% scrvUSD yield (APY) 7.9% +1.1 pp scrvUSD assets / mint debt 47.1% +12.1 pp crvUSD oracle price $0.9969 -$0.0007 Average borrow APY 7.8% +0.9 pp PegKeeper debt 0.00M crvUSD +0.00M crvUSD Mint borrowing fees collected $57.3k -16.0% Llamalend Metric Value Vs W46 🦙 Llamalend TVL $192M -15.2% 🦙 Supplied $108M -7.2% 🦙 Borrowed $102M -14.1% 🦙 Collateral $144M -19.0% 🦙 Loans 1,170 -8 🦙 V2 accrued admin fees $2.5k -2.8% TVL counts lending and mint markets; supplied assets are lending-only. DEX Metric Value Vs W46 🔄 Volume $826.8M -7.2% 🔄 Swaps 382k +29.7% 🔄 Trading fees $209.8k +16.2% Pools that earned the most trading fees this week: Chain Pool Volume Fees USDT WBTC WETH $6.4M $29.5k USDT WBTC WETH $41.4M $18.1k USDC fxUSD $142.3M $15.6k USDS stUSDS $7.4M $11.5k WETH CVX $3.4M $9.9k apxUSD USDC $4.0M $9.7k USDC WBTC WETH $2.4M $7.8k DAI USDC USDT $50.5M $7.6k ETH stETH $75.2M $7.5k crvUSD tBTC wstETH $7.9M $7.4k Trading fees are the fees traders paid for swaps. They are shared between liquidity providers and the DAO. DEX figures cover ten chains and exclude YieldBasis pools, which have their own fee model. DAO Metric Value Vs W46 veCRV distribution $35.6k -73.8% veCRV APR 1.217% -2.624 pp Vote incentives (2-week avg) $222.3k -1.6% veCRV APR with incentives 11.123% -1.008 pp CRV emissions (weekly gauge allocation) 1.86M CRV ($367.4k) 0.00% Annual CRV inflation 3.989% -0.003 pp Weekly CRV emissions are scheduled gauge allocations, not tokens already claimed. Their change is measured in CRV, not dollars. Vote incentives are paid on Votemarket and Votium and averaged over two weeks, as Votium pays once per two-week round. Their APR counts only veCRV that voted for gauges with incentives. Governance and protocol updates Two new Llamalend v2 markets are live Two new Llamalend v2 markets take wOUSD , Origin's wrapped dollar stablecoin, and wstGBP , a staked British pound stablecoin, as collateral. The wOUSD market can lend up to 950k crvUSD and the wstGBP market up to 80k. You can supply crvUSD to fund those loans. Oracle pricing for YieldBasis pools A vote would have the YieldBasis WETH, WBTC, cbBTC and tBTC pools set swap prices from Pyth market data . Traders who close a price gap would pay part of their profit to the pool. Vote closes November 21. New gauges The DAO added a gauge for the DOLA/SwanUSD pool , so its LPs can earn CRV. Votes on new gauges are open for yscvxCRV/CRV , MET/WETH and bdUSD/MUSD . These votes close November 20, 23 and 24, in that order. Explore Curve Explore Curve pools , Llamalend markets or savings crvUSD . Choose the product that fits the asset and liquidity you need. Risk information: pool risks , lending risks and crvUSD risks . Link

medium.com: Espresso’s 1 Million TPS Is No Longer Theoretical<br /> Espresso’s 1 Million TPS Is No Longer Theoretical Last month we said Espresso could finalize up to one million transactions per second. The number came from measured throughput of 181 MB/s divided by a 180 bytes, the average size of an ERC-20 transfer. The math was sound. But no one had actually watched the protocol finalize a million transactions per second without slowing down. Now we have. We spun up 87 independent chains, each with its own sequencer and execution client, and had them all flood Espresso with ERC-20 transfers at the same time. As you can see in the video below, Espresso finalized more than one million transactions per second while keeping average finality under one second. The 1M TPS claim is no longer theoretical. For institutions evaluating onchain infrastructure, this is the result that matters. A collateral engine, payments platform, and treasury desk can each run as their own application with their own rules, and every counterparty can treat a transaction as final the moment Espresso confirms it, even at peak volume. Here’s a video with Espresso Systems CTO Jeb Bearer explaining what his team built while showing one of the demo’s runs, followed by the full methodology and numbers. The results 1,000,000+ — transactions per second, sustained across 87 chains <0.5 seconds — Espresso finality (same as Espresso’s block time), holding steady from ramp-up through peak load ~168 MB/s — data finalized by Espresso every second (didn’t even need the full 181 MB/s) As the load generators ramped up, the transaction count climbed past a million per second and stayed there. Finality held in the same narrow band the whole time, at roughly a half a second, so adding load didn’t add wait time. What we ran Our engineers stood up 87 independent chains, each built with the Espresso Stack, a lean framework we developed that derives a chain’s state directly from Espresso consensus so the chain inherits Espresso’s security and performance. We first used the Espresso Stack in the crosschain collateral management demo we built with AWS and introduced onstage at the Digital Asset Summit in March. We used the Besu execution client and gave each chain its own sequencer and a dedicated load generator firing upwards of 12,000 ERC-20 token transfers at Espresso per second. Espresso’s job was to finalize the order of transactions and the data behind them as fast as possible without getting bogged down. The tests ran on a 100-node Espresso devnet built to mirror mainnet. While all nodes sat in one AWS region, we purposefully slowed down the messages passing between them to simulate real round-trip times between Europe, Southeast Asia, the US East Coast, Australia and South America, so consensus faced the same delays that Espresso mainnet’s globally distributed validator set would. To ensure this wasn’t a one-off result, we ran the tests dozens of times, including a 10-minute run. The result was always the same. Espresso consistently finalized more than 1 million transactions per second while maintaining subsecond latency. The full setup is in the methodology section below. Proof the data was really there Counting the number of transactions Espresso validators can finalize every second only matters if the data behind those transactions can be used to verify the correct state of an Espresso-integrated chain. To that end, we stood up a full node for one of the Besu chains, which we called a follower node, that had no connection to that chain’s sequencer and rebuilt the chain’s entire state using nothing but what Espresso finalized. Throughout the run, the follower stayed just a few blocks behind the sequencer. If Espresso had fallen behind or dropped data under load, that gap would have widened, but it never did. Any third party reading from Espresso, whether another chain, a bridge, an exchange or even an auditor, could have tracked that chain’s state in real time while the network was handling more than a million transactions per second. Why Espresso can do this Espresso was designed from day one to focus on consensus and finalizing transactions from independent chains as fast as possible, while leaving execution to those chains. Each Espresso-integrated chain orders and executes its own transactions through its own sequencer, and Espresso confirms that ordering, finalizes it and guarantees the data is available, which means total throughput grows as more chains connect while finality stays fast. That stripped-down design is what made two years of steady performance gains possible. The Espresso Stack is the native way to build chains for this design. Earlier integrations with chains running Arbitrum Nitro or the OP Stack needed workarounds, because those frameworks were built around a different model, whereas the Espresso Stack was written for Espresso from the start. Espresso Stack chains carry only what its job requires, with no general-purpose virtual machine, gas token or permissionless contract deployment to pay for, and crosschain tooling comes built in so any outside party can read a chain’s finalized state immediately. What this means for institutions The financial system of the future will generate far more transactions than blockchains handle today, and each one will get heavier. AI agents that buy, sell and rebalance on their own and at a pace impossible for humans will skyrocket transaction volume, while post-quantum signatures, compliance data and privacy proofs will add bytes to every transaction. That pressure lands on the property finance cares about most. A trade, a payment or a collateral call is only useful once it’s final, and a network that slows down under load forces every counterparty to wait. For a bank, an asset manager or a market infrastructure provider, the benchmark shows that finality can hold steady while volume and transaction size both climb. Institutions also don’t have to give anything up to get there. Each application can run as its own dedicated chain with its own rules, fees and permissions, without sharing an execution environment with anyone else. A settlement network that finalizes more than a million transfers per second leaves room for peak-day volume and for the heavier transactions ahead, and because finality arrives in under half a second, counterparties, custodians and other chains can trust a transaction the moment it’s final. Get in touch Whether you’re exploring institutional blockchain infrastructure or want to go deeper on the demo, we’d like to hear from you. Email us at hello@espressosys.com. If you’re a team that already runs a private, permissioned Besu network and are interested in exploring a move to an Espresso-based application environment, check out our docs. Methodology Devnet The Espresso devnet used in the demo ran as a 100-node network, with a handful of larger query nodes serving data and the rest running as lightweight nodes. Simulated round-trip latency (ms) To mirror Espresso mainnet’s globally distributed validator set, we added latency to simulate real-world delays. Every packet between two simulated regions carried the measured average round-trip time between those AWS regions. Packets within a region got no added delay, so real in-region latency applied. Chains and load We stood up 87 chains, each with its own execution client, sequencer and load generator sending transactions to Espresso. About Espresso Systems Espresso Systems builds digital asset infrastructure for institutions and enterprises that demand speed and customizability without sacrificing architectural control. It is lead developer of Espresso, a high-performance blockchain that provides the subsecond finality and massive data throughput making real-time settlement possible at institutional scale. Unlike legacy blockchains that impose their own rules on applications that build on them, Espresso provides a foundation for custom applications that need decentralized security and real-time finality without forcing them to sacrifice speed, connectivity, or architectural control. Companies and blockchains like Celo, ApeChain, Gate, Litecoin Foundation, and others have already chosen Espresso for the architectural control and native interoperability legacy infrastructure cannot match. Founded in 2020 by a team out of Stanford’s cryptography program, Espresso Systems has pushed the state of the art in blockchain, consensus protocol, and distributed network design. It is backed by over $50 million of capital from Andreessen Horowitz, Sequoia Capital, Greylock Partners, Coinbase Ventures, Electric Capital, and more. Website | Docs | GitHub | X | HackMD | LinkedIn | Espresso Foundation Link

medium.com: Inside the java-tron v4.8.3(Antisthenes) Admin: Online Node Management and Peer Operations<br /> Inside the java-tron v4.8.3(Antisthenes) Admin: Online Node Management and Peer Operations Java-tron 4.8.3 will introduce an Admin interface that provides interactive and scriptable management capabilities for a running FullNode. Node operators can inspect peer connections and synchronization status, adjust the active peer set, disconnect specific connections, and maintain a manual IP blocklist that persists across node restarts. These operations can be performed online without restarting the node for each adjustment. The first set of Admin features focuses on peer operations and provides two access methods: a local IPC console and a standalone HTTP service. This article begins with the overall architecture, explains the scope of each of the seven commands, and then walks through a connection-handling scenario covering status inspection, operation execution, and result verification. 1. Architecture: Two Access Paths, One JSON-RPC Interface Admin uses JSON-RPC to define methods, parameters, results, and errors consistently. IPC and HTTP serve only as different transport mechanisms. The same management command follows the same business rules regardless of which access path is used. Local CLI (--attach / --exec) ─→ IPC ─┐ ├─→ Admin JSON-RPC ─→ FullNode peer management Automation (POST /admin) ────────→ HTTP ─┘ The two access paths are designed for different scenarios: Local IPC: Intended for interactive troubleshooting after logging into the node host. Use FullNode.jar --attach to connect to an existing node. The client provides command help, completion, and formatted output, while --exec executes a single command and exits. Standalone HTTP: Intended for integration with curl or operational software, exposing POST /admin on a dedicated port. It is configured separately from the existing public JSON-RPC service and can be enabled or disabled independently of IPC. Importantly, --attach mode runs a client. It does not start a second FullNode or open the node database again. Admin’s design draws on the node management, IPC, and RPC access-control approaches used by Geth and Besu, while providing standalone HTTP and local IPC access for the java-tron runtime environment. For java-tron, the configuration and commands described in this article and the project documentation should be treated as authoritative. 2. Seven Commands: Queries, Active Peers, One-Off Disconnects, and IP Blocking The initial commands can be divided into four groups based on their target and scope: 2.1 Inspect Current Connections admin_listActivePeers: Displays connection and synchronization statistics for currently connected peers, including connection direction, connection duration, average latency, last known block height, synchronization queues, and inactivity duration. 2.2 Adjust the Runtime Active Peer Set admin_addPeer: Adds the specified IP address and port to the runtime active peer set. admin_removePeer: Removes the specified endpoint from the active peer set and disconnects its current connection. Adding and removing active peers uses the IPv4:port or [IPv6]:port format. A successful addition only means that the endpoint has entered the active peer set. The P2P module still attempts connections according to its normal strategy; success does not mean that a TCP connection or handshake has completed, and it does not automatically add the peer to the trusted node set. Active peers added through Admin remain effective only for the current process lifetime. 2.3 Disconnect a Specific Connection Once admin_disconnectPeer: Disconnects the current connection for the specified IP address and port while preserving the active peer set. A one-off disconnect is not a permanent removal. If the peer remains in the active peer set or is connected again later, it may reappear in the connection list. 2.4 Maintain the Manual IP Blocklist admin_blockIp: Persistently blocks the specified IP address and disconnects its existing TCP connections. admin_unblockIp: Removes the specified IP address from the manual IP blocklist. admin_listBlockedIps: Lists manually blocked IP addresses and the times when their records were created. Unlike runtime active peers, manual IP block records are restored after a node restart and take precedence over the active peer connection strategy. Blocking the same IP address again is a successful no-op and preserves the original record timestamp. After an address is unblocked, whether a connection can be re-established still depends on normal connection conditions. Manual blocking accepts IPv4 or IPv6 addresses and applies to TCP connections and dialing candidates, but not UDP discovery traffic. It currently does not support domain names, CIDR ranges, or node IDs, and it provides neither automatic expiration nor automatic malicious-node detection. Blocking decisions should therefore be based on logs, connection state, and operational judgment. 2.5 Restrictions in Dynamic Configuration Mode If node.dynamicConfig.enable is enabled, active peers remain managed through node.active in the configuration file. In this mode, admin_addPeer and admin_removePeer return a business failure and direct the operator to modify the configuration. One-off disconnects, connection queries, and manual IP blocking remain available. 3. Getting Started: Connect to a Node Through Local IPC Both Admin access paths are disabled by default. For initial use, you can enable only local IPC by merging the following settings into the node configuration file and starting FullNode with them: node.admin.ipc { enable = true socketDirectory = "" } When socketDirectory is empty, the socket is placed in the ipc subdirectory of the node output directory, and its filename includes the node process PID. For example, run the following command from /data/fullnode: java -jar FullNode.jar \ --attach output-directory/ipc/2355169.sock In practice, replace the path and PID with the current node’s socket path. After entering the console, use the following commands for help: help help admin_disconnectPeer 4. Practical Example: Disconnecting a Long-Inactive Connection Suppose a node experiences intermittent synchronization stalls, and an operator finds a connection that has had no block-related activity for a long time. Rather than immediately blocking the address, the appropriate sequence is to inspect its synchronization and task state first, and then decide whether a one-off disconnect is warranted. 4.1 Inspect Connections and Select a Target > admin_listActivePeers text Peers: all=50, active=3, passive=47, effective=49 This query returns 50 connections: 3 initiated locally and 47 accepted inbound. Here, activeand passive describe the connection direction; they are not the same concept as the active peer set in the configuration. The following excerpt shows selected connections and fields, sorted by Idle in descending order: Address Idle (s) Block Sync P/U 192.0.2.10:20453 562222 0 N/N 192.0.2.20:33649 562212 0 N/N 192.0.2.30:51185 561720 0 N/N 198.51.100.10:38721 0 88104889 N/N All peer IP addresses in this article are documentation examples. In a real operation, use the addresses returned by the current node. Idle corresponds to inactiveSeconds, the number of seconds since the most recent block-related activity recorded by the node. It can help identify long-inactive connections, but: A high Idle value does not imply malicious behavior. A high Idle value does not by itself prove that the connection is faulty. The field does not mean that no data has been sent or received over the TCP connection. This example selects 192.0.2.10:20453, which has the highest Idle value: 562,222 seconds, or approximately 6.5 days. The complete query also shows Sync P/U as N/N, Req/Remain as 0/0, and Proc as 0, with no pending synchronization requests or blocks being processed. Based on this information, the operator decides to disconnect it once rather than impose a long-term block. 4.2 Execute the Disconnect and Inspect the Return Values > admin_disconnectPeer 192.0.2.10:20453 { "success" : true, "changed" : false, "disconnectedCount" : 1, "errorMessage" : "" } These fields describe three different facts: success=true: The management operation succeeded. disconnectedCount=1: One connection was disconnected. changed=false: The active peer set was not modified. A one-off disconnect preserves that set by design, so this value does not indicate failure. 4.3 Query Again to Verify Connection State After running admin_listActivePeers text again, the target address has disappeared from the connection list. A new address, 203.0.113.10:19471, appears with a connection duration of 9 seconds, while the total connection count remains 50. In this example, the disappearance of the target address and the returned disconnect count confirm that the operation was executed. The effective count is 49 in both queries, so this record only demonstrates online inspection and intervention. It does not demonstrate an improvement in the effective connection count or synchronization performance. 5. Integrating Operational Workflows Through Scripts and HTTP 5.1 Execute a Single Command with --exec Console queries can also be executed as single commands, allowing inspection scripts to collect current connection state: java -jar FullNode.jar \ --attach output-directory/ipc/2355169.sock \ --exec "admin_listActivePeers text" A successful one-shot execution returns exit code 0. A command error, JSON-RPC error, peer operation returning success=false, or communication failure returns a non-zero exit code. Scripts can use this behavior to handle success and failure and apply a timeout to the entire command as needed. 5.2 Enable an HTTP Endpoint Bound to Loopback The HTTP endpoint is suitable for automation, but Admin HTTP currently has no built-in authentication. The following configuration keeps the listener on the local loopback address: node.admin.http { enable = true listenAddress = "127.0.0.1" port = 8575 virtualHosts = ["localhost"] } The HTTP equivalent of the preceding connection query is: curl -sS http://127.0.0.1:8575/admin \ -H 'Content-Type: application/json' \ --data '{"jsonrpc":"2.0","id":1,"method":"admin_listActivePeers","params":[]}' HTTP returns a complete JSON-RPC response, while the text table format is generated by the IPC client. Admin currently accepts only single JSON-RPC requests and does not support batch requests. 5.3 Do Not Treat HTTP 200 as Proof of Operational Success HTTP callers should inspect the error field in the JSON-RPC response. For modifying operations, they should also inspect business fields such as result.success, changed, and disconnectedCount. HTTP 200 only means that the HTTP request received a response; it does not prove that the management operation succeeded. Automation must handle transport errors, JSON-RPC errors, and business failures separately. 6. Enable Access Selectively and Protect the Management Endpoints Admin is intended for FullNode operators, and both HTTP and IPC are disabled by default. Enabling an existing public JSON-RPC service does not automatically expose these management commands. 6.1 HTTP: Neither Listen Addresses nor Host Checks Replace Access Control listenAddress = 0.0.0.0 listens on all IPv4 network interfaces. If the port is publicly reachable and allowed by the firewall or security group, the Admin port may be exposed to the entire internet. Because the service has no built-in authentication, the default 127.0.0.1listener is recommended. If remote access is required, restrict it through deployment controls such as a managed network, firewall rules, reverse-proxy authentication, or mTLS. virtualHosts only validates the HTTP Host header; it does not replace user authentication or client source-IP access control. 6.2 IPC: Control Socket Directories and File Permissions IPC uses private directories and socket-file permissions to restrict local access. Deployments should meet the following requirements: The socket root directory should be controlled by the node user or a trusted administrator and located on a filesystem that supports POSIX permissions. A custom socketDirectory must be an existing absolute directory. The complete socket path must be no more than 100 encoded bytes. Each node should use a separate socket root directory. An abnormal exit may leave an ipc directory behind and cause the next startup to fail. Before cleanup, confirm that no node is using the directory. Then follow the configuration documentation to remove only verified stale socket files and empty directories, avoiding IPC files that are still in use. 7. Operational Rollout: Start with Read-Only Queries After upgrading, start by enabling local IPC and using read-only queries such as admin_listActivePeers. Become familiar with connection direction, Idle, synchronization queues, and return values before introducing modifying operations: Inspect before acting: Evaluate targets using connection state, logs, and synchronization tasks rather than a single metric. Choose the correct scope: Use admin_disconnectPeer for a one-off disconnect, the add/remove commands for the runtime active peer set, and manual blocking only when an IP address must remain blocked across restarts. Verify the actual result: Check both return values and subsequent queries rather than relying only on command submission or HTTP 200. Limit endpoint exposure: Keep unused access paths disabled, prefer loopback for HTTP, and use controlled private directories for IPC. Automate afterward: Handle non-zero exit codes, JSON-RPC errors, and business failures in scripts, and set appropriate command timeouts. The first Admin capabilities in java-tron 4.8.3 move peer operations beyond observing state toward executing changes online and verifying their results. A unified interface, well-defined operation semantics, and access paths that are disabled by default provide a common foundation for both interactive troubleshooting and automated operations. 8. Further Reading Admin management framework design, Issue #6497 Peer management and manual IP blocking, Issue #6989 Admin HTTP and IPC implementation, PR #6979 Peer management implementation, PR #7035 Admin configuration and IPC recovery guide Geth Admin documentation Besu JSON-RPC authentication and authorization documentation Link

medium.com: Inside the java-tron v4.8.3(Antisthenes) Admin: Online Node Management and Peer Operations<br /> Inside the java-tron v4.8.3(Antisthenes) Admin: Online Node Management and Peer Operations Java-tron 4.8.3 will introduce an Admin interface that provides interactive and scriptable management capabilities for a running FullNode. Node operators can inspect peer connections and synchronization status, adjust the active peer set, disconnect specific connections, and maintain a manual IP blocklist that persists across node restarts. These operations can be performed online without restarting the node for each adjustment. The first set of Admin features focuses on peer operations and provides two access methods: a local IPC console and a standalone HTTP service. This article begins with the overall architecture, explains the scope of each of the seven commands, and then walks through a connection-handling scenario covering status inspection, operation execution, and result verification. 1. Architecture: Two Access Paths, One JSON-RPC Interface Admin uses JSON-RPC to define methods, parameters, results, and errors consistently. IPC and HTTP serve only as different transport mechanisms. The same management command follows the same business rules regardless of which access path is used. Local CLI (--attach / --exec) ─→ IPC ─┐ ├─→ Admin JSON-RPC ─→ FullNode peer management Automation (POST /admin) ────────→ HTTP ─┘ The two access paths are designed for different scenarios: Local IPC: Intended for interactive troubleshooting after logging into the node host. Use FullNode.jar --attach to connect to an existing node. The client provides command help, completion, and formatted output, while --exec executes a single command and exits. Standalone HTTP: Intended for integration with curl or operational software, exposing POST /admin on a dedicated port. It is configured separately from the existing public JSON-RPC service and can be enabled or disabled independently of IPC. Importantly, --attach mode runs a client. It does not start a second FullNode or open the node database again. Admin’s design draws on the node management, IPC, and RPC access-control approaches used by Geth and Besu, while providing standalone HTTP and local IPC access for the java-tron runtime environment. For java-tron, the configuration and commands described in this article and the project documentation should be treated as authoritative. 2. Seven Commands: Queries, Active Peers, One-Off Disconnects, and IP Blocking The initial commands can be divided into four groups based on their target and scope: 2.1 Inspect Current Connections admin_listActivePeers: Displays connection and synchronization statistics for currently connected peers, including connection direction, connection duration, average latency, last known block height, synchronization queues, and inactivity duration. 2.2 Adjust the Runtime Active Peer Set admin_addPeer: Adds the specified IP address and port to the runtime active peer set. admin_removePeer: Removes the specified endpoint from the active peer set and disconnects its current connection. Adding and removing active peers uses the IPv4:port or [IPv6]:port format. A successful addition only means that the endpoint has entered the active peer set. The P2P module still attempts connections according to its normal strategy; success does not mean that a TCP connection or handshake has completed, and it does not automatically add the peer to the trusted node set. Active peers added through Admin remain effective only for the current process lifetime. 2.3 Disconnect a Specific Connection Once admin_disconnectPeer: Disconnects the current connection for the specified IP address and port while preserving the active peer set. A one-off disconnect is not a permanent removal. If the peer remains in the active peer set or is connected again later, it may reappear in the connection list. 2.4 Maintain the Manual IP Blocklist admin_blockIp: Persistently blocks the specified IP address and disconnects its existing TCP connections. admin_unblockIp: Removes the specified IP address from the manual IP blocklist. admin_listBlockedIps: Lists manually blocked IP addresses and the times when their records were created. Unlike runtime active peers, manual IP block records are restored after a node restart and take precedence over the active peer connection strategy. Blocking the same IP address again is a successful no-op and preserves the original record timestamp. After an address is unblocked, whether a connection can be re-established still depends on normal connection conditions. Manual blocking accepts IPv4 or IPv6 addresses and applies to TCP connections and dialing candidates, but not UDP discovery traffic. It currently does not support domain names, CIDR ranges, or node IDs, and it provides neither automatic expiration nor automatic malicious-node detection. Blocking decisions should therefore be based on logs, connection state, and operational judgment. 2.5 Restrictions in Dynamic Configuration Mode If node.dynamicConfig.enable is enabled, active peers remain managed through node.active in the configuration file. In this mode, admin_addPeer and admin_removePeer return a business failure and direct the operator to modify the configuration. One-off disconnects, connection queries, and manual IP blocking remain available. 3. Getting Started: Connect to a Node Through Local IPC Both Admin access paths are disabled by default. For initial use, you can enable only local IPC by merging the following settings into the node configuration file and starting FullNode with them: node.admin.ipc { enable = true socketDirectory = "" } When socketDirectory is empty, the socket is placed in the ipc subdirectory of the node output directory, and its filename includes the node process PID. For example, run the following command from /data/fullnode: java -jar FullNode.jar \ --attach output-directory/ipc/2355169.sock In practice, replace the path and PID with the current node’s socket path. After entering the console, use the following commands for help: help help admin_disconnectPeer 4. Practical Example: Disconnecting a Long-Inactive Connection Suppose a node experiences intermittent synchronization stalls, and an operator finds a connection that has had no block-related activity for a long time. Rather than immediately blocking the address, the appropriate sequence is to inspect its synchronization and task state first, and then decide whether a one-off disconnect is warranted. 4.1 Inspect Connections and Select a Target > admin_listActivePeers text Peers: all=50, active=3, passive=47, effective=49 This query returns 50 connections: 3 initiated locally and 47 accepted inbound. Here, activeand passive describe the connection direction; they are not the same concept as the active peer set in the configuration. The following excerpt shows selected connections and fields, sorted by Idle in descending order: Address Idle (s) Block Sync P/U 192.0.2.10:20453 562222 0 N/N 192.0.2.20:33649 562212 0 N/N 192.0.2.30:51185 561720 0 N/N 198.51.100.10:38721 0 88104889 N/N All peer IP addresses in this article are documentation examples. In a real operation, use the addresses returned by the current node. Idle corresponds to inactiveSeconds, the number of seconds since the most recent block-related activity recorded by the node. It can help identify long-inactive connections, but: A high Idle value does not imply malicious behavior. A high Idle value does not by itself prove that the connection is faulty. The field does not mean that no data has been sent or received over the TCP connection. This example selects 192.0.2.10:20453, which has the highest Idle value: 562,222 seconds, or approximately 6.5 days. The complete query also shows Sync P/U as N/N, Req/Remain as 0/0, and Proc as 0, with no pending synchronization requests or blocks being processed. Based on this information, the operator decides to disconnect it once rather than impose a long-term block. 4.2 Execute the Disconnect and Inspect the Return Values > admin_disconnectPeer 192.0.2.10:20453 { "success" : true, "changed" : false, "disconnectedCount" : 1, "errorMessage" : "" } These fields describe three different facts: success=true: The management operation succeeded. disconnectedCount=1: One connection was disconnected. changed=false: The active peer set was not modified. A one-off disconnect preserves that set by design, so this value does not indicate failure. 4.3 Query Again to Verify Connection State After running admin_listActivePeers text again, the target address has disappeared from the connection list. A new address, 203.0.113.10:19471, appears with a connection duration of 9 seconds, while the total connection count remains 50. In this example, the disappearance of the target address and the returned disconnect count confirm that the operation was executed. The effective count is 49 in both queries, so this record only demonstrates online inspection and intervention. It does not demonstrate an improvement in the effective connection count or synchronization performance. 5. Integrating Operational Workflows Through Scripts and HTTP 5.1 Execute a Single Command with --exec Console queries can also be executed as single commands, allowing inspection scripts to collect current connection state: java -jar FullNode.jar \ --attach output-directory/ipc/2355169.sock \ --exec "admin_listActivePeers text" A successful one-shot execution returns exit code 0. A command error, JSON-RPC error, peer operation returning success=false, or communication failure returns a non-zero exit code. Scripts can use this behavior to handle success and failure and apply a timeout to the entire command as needed. 5.2 Enable an HTTP Endpoint Bound to Loopback The HTTP endpoint is suitable for automation, but Admin HTTP currently has no built-in authentication. The following configuration keeps the listener on the local loopback address: node.admin.http { enable = true listenAddress = "127.0.0.1" port = 8575 virtualHosts = ["localhost"] } The HTTP equivalent of the preceding connection query is: curl -sS http://127.0.0.1:8575/admin \ -H 'Content-Type: application/json' \ --data '{"jsonrpc":"2.0","id":1,"method":"admin_listActivePeers","params":[]}' HTTP returns a complete JSON-RPC response, while the text table format is generated by the IPC client. Admin currently accepts only single JSON-RPC requests and does not support batch requests. 5.3 Do Not Treat HTTP 200 as Proof of Operational Success HTTP callers should inspect the error field in the JSON-RPC response. For modifying operations, they should also inspect business fields such as result.success, changed, and disconnectedCount. HTTP 200 only means that the HTTP request received a response; it does not prove that the management operation succeeded. Automation must handle transport errors, JSON-RPC errors, and business failures separately. 6. Enable Access Selectively and Protect the Management Endpoints Admin is intended for FullNode operators, and both HTTP and IPC are disabled by default. Enabling an existing public JSON-RPC service does not automatically expose these management commands. 6.1 HTTP: Neither Listen Addresses nor Host Checks Replace Access Control listenAddress = 0.0.0.0 listens on all IPv4 network interfaces. If the port is publicly reachable and allowed by the firewall or security group, the Admin port may be exposed to the entire internet. Because the service has no built-in authentication, the default 127.0.0.1listener is recommended. If remote access is required, restrict it through deployment controls such as a managed network, firewall rules, reverse-proxy authentication, or mTLS. virtualHosts only validates the HTTP Host header; it does not replace user authentication or client source-IP access control. 6.2 IPC: Control Socket Directories and File Permissions IPC uses private directories and socket-file permissions to restrict local access. Deployments should meet the following requirements: The socket root directory should be controlled by the node user or a trusted administrator and located on a filesystem that supports POSIX permissions. A custom socketDirectory must be an existing absolute directory. The complete socket path must be no more than 100 encoded bytes. Each node should use a separate socket root directory. An abnormal exit may leave an ipc directory behind and cause the next startup to fail. Before cleanup, confirm that no node is using the directory. Then follow the configuration documentation to remove only verified stale socket files and empty directories, avoiding IPC files that are still in use. 7. Operational Rollout: Start with Read-Only Queries After upgrading, start by enabling local IPC and using read-only queries such as admin_listActivePeers. Become familiar with connection direction, Idle, synchronization queues, and return values before introducing modifying operations: Inspect before acting: Evaluate targets using connection state, logs, and synchronization tasks rather than a single metric. Choose the correct scope: Use admin_disconnectPeer for a one-off disconnect, the add/remove commands for the runtime active peer set, and manual blocking only when an IP address must remain blocked across restarts. Verify the actual result: Check both return values and subsequent queries rather than relying only on command submission or HTTP 200. Limit endpoint exposure: Keep unused access paths disabled, prefer loopback for HTTP, and use controlled private directories for IPC. Automate afterward: Handle non-zero exit codes, JSON-RPC errors, and business failures in scripts, and set appropriate command timeouts. The first Admin capabilities in java-tron 4.8.3 move peer operations beyond observing state toward executing changes online and verifying their results. A unified interface, well-defined operation semantics, and access paths that are disabled by default provide a common foundation for both interactive troubleshooting and automated operations. 8. Further Reading Admin management framework design, Issue #6497 Peer management and manual IP blocking, Issue #6989 Admin HTTP and IPC implementation, PR #6979 Peer management implementation, PR #7035 Admin configuration and IPC recovery guide Geth Admin documentation Besu JSON-RPC authentication and authorization documentation Link

medium.com: 100xSOON: Cara AI Agent Trading di AI Capital Markets SOON Network<br /> 100xSOON: Cara AI Agent Trading di AI Capital Markets SOON Network 100xSOON adalah platform trading leverage tinggi kami untuk AI agent keuangan. Platform ini berjalan di BNB Chain dan juga terbuka untuk trader manusia. AI agent (agen AI) terhubung lewat MCP, membayar lewat x402, dan bertransaksi dalam $SOON. Jaringan: BNB Chain Pasar: BTC, ETH, SOL; NVDA, TSLA, AAPL, CRCL; emas dan perak Leverage maksimum: Hingga 10.000x untuk BTC dan hingga 5.000x untuk saham Penyelesaian: Otomatis, 30 detik setelah posisi dibuka Akses agent: Model Context Protocol (MCP) Metode pembayaran: Protokol x402 Token trading: Hanya $SOON; kontrak BNB Chain 0xb9E1Fd5A02D3A33b25a14d661414E6ED6954a721 Situs: https://www.100xsoon.com/ Posisi 100xSOON dalam AI Capital Markets AI Capital Markets (pasar modal AI) adalah infrastruktur on-chain yang kami bangun di SOON agar AI agent dapat saling bertransaksi, membayar, dan berkoordinasi. 100xSOON adalah bagian trading-nya. Platform ini memberi agent tempat untuk mengambil posisi pada kripto, saham, dan logam tanpa ada orang yang memasang order. Kami pertama kali meluncurkan 100xSOON di Base pada Januari 2026 sebagai produk trading perpetual berbasis x402. Pada September 2026 kami meluncurkan versi BNB Chain dengan akses agent. Apa yang dapat diperdagangkan AI agent 100xSOON menyediakan tiga kelompok pasar: aset kripto (BTC, ETH, SOL), saham AS (NVDA, TSLA, AAPL, CRCL), dan logam mulia (emas dan perak). Kami akan menambahkan saham lain. Leverage berbeda untuk tiap pasar. Batas maksimumnya 10.000x untuk BTC dan 5.000x untuk saham. Pilihan margin bergantung pada leverage yang dipilih. Pada 10.000x, margin maksimum adalah 5 $SOON, yang membuka posisi senilai 50.000 $SOON. Tidak ada syarat khusus untuk memakai tingkat leverage tinggi. Setelah posisi dibuka, hitung mundur dimulai, dan posisi diselesaikan secara otomatis setelah 30 detik. Kami dapat menyesuaikan biaya, pilihan margin, dan parameter trading lainnya. Nilai yang berlaku selalu ditampilkan di aplikasi dan di dokumentasi. Cara AI agent terhubung: MCP dan x402 Trading AI agent di 100xSOON memakai dua standar terbuka. Model Context Protocol (MCP) adalah standar terbuka untuk menghubungkan aplikasi AI dengan sistem eksternal. Anthropic memperkenalkannya pada akhir 2024 dan menyerahkannya kepada Agentic AI Foundation di bawah Linux Foundation pada Desember 2025. x402 adalah standar pembayaran terbuka yang dibuat Coinbase dan diluncurkan pada Mei 2025. Standar ini menempatkan permintaan pembayaran di dalam permintaan web biasa, sehingga perangkat lunak dapat membayar tanpa langkah checkout manual. 100xSOON terhubung dengan agent lewat MCP. Berikan dokumentasi kami di https://www.100xsoon.com/skill.md kepada agent Anda; dokumentasi itu menjelaskan cara terhubung. Setelah itu, agent dapat membaca data pasar, membuka posisi, dan bertransaksi sendiri. Claude, Codex, Hermes, dan agent lain yang kompatibel dengan MCP dapat terhubung. Pembayaran berjalan lewat x402, sehingga agent membayar langsung dari wallet miliknya. Kami tidak pernah meminta private key atau seed phrase. Permintaan atas salah satunya adalah penipuan. Peran token SOON Semua trading di 100xSOON memakai denominasi $SOON, dan margin serta pembayaran memakai token yang sama. Kami memakai 80% pendapatan platform untuk buyback dan burn $SOON. Pendapatan diterima dalam bentuk token, sehingga nilainya dalam dolar berubah mengikuti harga token, dan jumlah yang dibakar bergantung pada aktivitas trading. Cara memulai trading AI agent atau trading manual Untuk AI agent Berikan dokumentasi di https://www.100xsoon.com/skill.md kepada agent. Isi wallet agent dengan $SOON di BNB Chain. Biarkan agent mulai bertransaksi dalam batas yang Anda tetapkan. Untuk trader manusia Hubungkan wallet di https://www.100xsoon.com/. Siapkan $SOON di BNB Chain. Token ini tersedia di PancakeSwap; periksa alamat kontrak pada tabel di atas sebelum melakukan swap. Pilih pasar, leverage, dan margin, lalu buka posisi. Risiko Trading leverage tinggi memiliki risiko kerugian yang tinggi. Seluruh margin dapat hilang dalam 30 detik, dan pada leverage yang sangat tinggi, pergerakan harga yang kecil sudah cukup untuk menyebabkannya. Biaya dipotong dari pembayaran. Periksa aturan biaya yang berlaku sebelum membuka posisi. Margin, keuntungan, dan kerugian seluruhnya dalam denominasi satu token, sehingga volatilitas harganya menambah risiko leverage. Dalam trading AI agent, agent dapat mengulang kesalahan yang sama berkali-kali dalam waktu singkat. Tetapkan batas ukuran posisi dan total dana. Ketersediaan dapat dibatasi di sebagian yurisdiksi. Periksa peraturan setempat sebelum bertransaksi. Disclaimer risiko investasi kripto: artikel ini hanya bersifat informasi dan bukan saran investasi. Gunakan hanya dana yang siap Anda tanggung kerugiannya. Referensi Dokumentasi agent 100xSOON: https://www.100xsoon.com/skill.md The Block, “What is Coinbase’s x402 protocol?”: https://www.theblock.co/learn/web3/what-is-coinbases-x402-protocol-391983 Anthropic, “Donating the Model Context Protocol and establishing the Agentic AI Foundation” (9 Desember 2025): https://anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation Link

medium.com: 100xSOON: How AI Agents Trade in SOON’s AI Capital Markets<br /> 100xSOON: How AI Agents Trade in SOON’s AI Capital Markets 100xSOON is our high-leverage trading platform for financial AI agents. It runs on BNB Chain and is also open to human traders. AI agents connect to it through MCP, pay through x402, and trade in $SOON. Chain: BNB Chain Markets: BTC, ETH, SOL; NVDA, TSLA, AAPL, CRCL; gold and silver Maximum leverage: Up to 10,000x on BTC and up to 5,000x on stocks Settlement: Automatic, 30 seconds after a position opens Agent access: Model Context Protocol (MCP) Payment method: x402 protocol Trading token: $SOON only; BNB Chain contract 0xb9E1Fd5A02D3A33b25a14d661414E6ED6954a721 Website: https://www.100xsoon.com/ Where 100xSOON sits in AI Capital Markets AI Capital Markets is the on-chain infrastructure we are building at SOON for AI agents to trade, pay, and coordinate with each other. 100xSOON is the trading part of it. It gives an agent a venue where it can take positions on crypto, stocks, and metals without a person placing the orders. We first launched 100xSOON on Base in January 2026 as an x402-based perpetual trading product. In September 2026 we launched the BNB Chain version with agent access. What AI agents can trade 100xSOON lists three groups of markets: crypto assets (BTC, ETH, SOL), US stocks (NVDA, TSLA, AAPL, CRCL), and precious metals (gold and silver). We will add more stocks. Leverage varies by market. The maximum is 10,000x on BTC and 5,000x on stocks. The margin options depend on the leverage selected. At 10,000x the maximum margin is 5 $SOON, which opens a position of 50,000 $SOON. There are no entry requirements for the high-leverage tiers. Once a position is open, a countdown starts, and the position settles automatically after 30 seconds. We may adjust fees, margin options, and other trading parameters. The current values are always shown in the app and in the documentation. How AI agents connect: MCP and x402 AI agent trading on 100xSOON relies on two open standards. Model Context Protocol (MCP) is an open standard for connecting AI applications to external systems. Anthropic introduced it in late 2024 and donated it to the Agentic AI Foundation under the Linux Foundation in December 2025. x402 is an open payment standard created by Coinbase and launched in May 2025. It places the payment request inside a normal web request, so that software can pay without a manual checkout step. 100xSOON connects to agents through MCP. Give your agent our documentation at https://www.100xsoon.com/skill.md, which explains how to connect. After that, the agent can read market data, open positions, and trade on its own. Claude, Codex, Hermes, and other MCP-compatible agents can connect. Payments go through x402, so the agent pays directly from its own wallet. We never ask for a private key or seed phrase. A request for either one is a scam. The role of $SOON All trading on 100xSOON is denominated in $SOON, and margin and payouts use the same token. We use 80% of platform revenue to buy back and burn $SOON. Revenue is collected in the token, so its value in dollar terms changes with the token price, and the amount burned depends on trading activity. How to start AI agent trading or manual trading For AI agents Give the agent the documentation at https://www.100xsoon.com/skill.md. Fund the agent’s wallet with $SOON on BNB Chain. Let the agent start trading within the limits you set. For human traders Connect a wallet at https://www.100xsoon.com/. Hold $SOON on BNB Chain. The token is listed on PancakeSwap; check the contract address in the table above before swapping. Choose a market, leverage, and margin, then open a position. Risks High-leverage trading carries a high risk of loss. The entire margin can be lost within 30 seconds, and at very high leverage a small price move is enough. Fees are deducted from payouts. Check the current fee rules before opening a position. Margin, profit, and loss are all denominated in one token, so its price volatility adds to the leverage risk. In AI agent trading, an agent can repeat an error many times in a short period. Set limits on position size and total funds. Availability may be restricted in some jurisdictions. Check local regulations before trading. This article is for information only and is not investment advice. Only trade what you can afford to lose. References 100xSOON agent documentation: https://www.100xsoon.com/skill.md The Block, “What is Coinbase’s x402 protocol?”: https://www.theblock.co/learn/web3/what-is-coinbases-x402-protocol-391983 Anthropic, “Donating the Model Context Protocol and establishing the Agentic AI Foundation” (December 9, 2025): https://anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation Link

medium.com: Privasi AI Agent di AI Capital Markets<br /> Privasi AI Agent di AI Capital Markets Pada September 2026 kami berinvestasi pada infrastruktur komputasi Phala Network dan menjadi investor utamanya. Kami melakukannya untuk menyelesaikan satu masalah yang spesifik. AI agent (agen AI) yang memegang private key dan mengelola dana biasanya berjalan di mesin cloud bersama, dan operator hardware memiliki sarana teknis untuk melihat model, strategi, dan kuncinya. Karena itu, privasi AI agent adalah syarat bagi AI Capital Markets (pasar modal AI), yaitu infrastruktur on-chain yang kami bangun di SOON agar agent dapat saling bertransaksi, membayar, dan berkoordinasi. Apa yang harus dijaga AI agent yang mengelola dana Agent yang bertransaksi dan membayar sendiri menangani beberapa jenis informasi sensitif: private key yang mengendalikan dananya; bobot model dan data yang menentukan kinerjanya; strategi yang dipakai untuk trading dan logika yang dipakai untuk penentuan harga; pihak lawan transaksi, anggaran, dan rencana upgrade; data pengguna dan agent lain yang berinteraksi dengannya. Dari mana paparan itu berasal Sebagian besar model AI berjalan di mesin cloud bersama. Pihak yang mengoperasikan hardware memiliki sarana teknis untuk memeriksa beban kerja di atasnya. Prompt, konteks, dan output melewati sistem yang tidak dikendalikan agent. Bagi asisten percakapan, dampak paparan ini terbatas. Bagi agent yang memegang kunci dan mengelola modal, dampaknya langsung: logika yang terlihat dapat disalin; strategi yang terlihat dapat didahului pihak lain (front-running); kunci yang terlihat memungkinkan aset dipindahkan keluar. Karena itu, privasi AI agent adalah syarat keamanan bagi setiap agent yang memegang aset. Cara kerja komputasi Phala TEE Trusted Execution Environment (TEE) adalah area pada prosesor yang diisolasi oleh hardware. Kode dan data di dalamnya dienkripsi di memori dan tidak dapat dibaca oleh sistem operasi host, hypervisor, maupun administrator mesin. TEE juga mendukung remote attestation. Hardware menghasilkan laporan bertanda tangan yang menyatakan kode dan firmware apa yang sedang berjalan. Pihak ketiga dapat memeriksa laporan itu dan memastikan apa yang dieksekusi tanpa melihat data yang sedang diproses. Komputasi bersifat privat dan sekaligus dapat diverifikasi. Komputasi Phala TEE untuk GPU disusun sebagai berikut: GPU: NVIDIA H100 dan H200 dalam mode confidential computing Sisi CPU: Confidential virtual machine di Intel TDX atau AMD SEV-SNP Atestasi: Laporan GPU diverifikasi melalui NVIDIA Remote Attestation Service; identitas dan kunci TEE dicatat on-chain Data yang dilindungi: Input, bobot model, dan output selama dipakai di GPU Benchmark Phala pada hardware H100 dan H200 menunjukkan rata-rata penurunan kinerja di bawah 9%. Mengapa agent keuangan di AI Capital Markets lebih dulu memerlukan privasi Aktivitas keuangan sedang berpindah ke on-chain. Robinhood Chain aktif di mainnet pada 1 Juli 2026 dengan token saham yang diperdagangkan 24 jam sehari, dan Robinhood meluncurkan agentic trading untuk saham dan opsi AS pada Juni 2026. Agent yang melakukan trading, hedging, rebalancing, dan manajemen risiko bergantung pada kerahasiaan. Strategi trading, posisi, pihak lawan transaksi, dan model milik sendiri tidak dapat beroperasi dalam kondisi terbuka bagi publik. AI Capital Markets dari SOON sudah mencakup 100xSOON, platform trading kami untuk agent. Komputasi Phala TEE menambahkan kerahasiaan yang diperlukan agent keuangan tersebut. Di SOON, kami juga berencana agar AI agent dapat meluncurkan dan mengoperasikan chain miliknya sendiri, dan kami akan mengalokasikan sebagian komputasi Phala TEE kami kepada agent tersebut. Batasan privasi AI agent berbasis TEE Kepercayaan pada hardware. TEE memindahkan kepercayaan dari operator cloud ke vendor hardware (NVIDIA, Intel, AMD) dan layanan atestasi mereka. Serangan fisik. Pada Oktober 2025, peneliti memublikasikan TEE.fail, serangan terhadap Intel TDX dan AMD SEV-SNP yang memakai perangkat yang dipasang pada bus memori DDR5. Serangan ini memerlukan akses fisik ke server. Cacat perangkat lunak. Atestasi memastikan kode mana yang sedang berjalan. Atestasi tidak memastikan bahwa kode itu bebas dari kesalahan. Transaksi publik. TEE melindungi komputasi. Transaksi yang dikirim agent ke chain publik tetap terlihat di chain tersebut. Disclaimer risiko investasi kripto: artikel ini hanya bersifat informasi dan bukan saran investasi. Referensi Phala, “GPU TEE Deep Dive: Securing AI at the Hardware Layer”: https://phala.com/posts/Phala-GPU-TEE-Deep-Dive The Block, peluncuran mainnet Robinhood Chain (1 Juli 2026): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading The Hacker News, serangan TEE.fail (Oktober 2025): https://thehackernews.com/2025/10/new-teefail-side-channel-attack.html Link

medium.com: AI Capital Markets의 AI 에이전트 프라이버시<br /> AI Capital Markets의 AI 에이전트 프라이버시 2026년 9월 저희는 팔라 네트워크(Phala Network)의 컴퓨팅 인프라에 투자했고, 주요 투자자가 되었습니다. 구체적인 문제 하나를 해결하기 위한 투자입니다. 개인 키를 보유하고 자금을 관리하는 AI 에이전트는 보통 공유 클라우드 장비에서 실행되며, 하드웨어 운영자는 에이전트의 모델, 전략, 키를 볼 수 있는 기술적 수단을 갖고 있습니다. 따라서 AI 에이전트 프라이버시는 AI Capital Markets(AI 자본시장)의 필수 요건입니다. AI Capital Markets는 에이전트가 서로 거래하고, 결제하고, 협업하도록 저희 쑨(SOON)이 구축하고 있는 온체인 인프라입니다. 자금을 관리하는 AI 에이전트가 지켜야 할 정보 스스로 거래하고 결제하는 에이전트는 여러 종류의 민감한 정보를 다룹니다. 자금을 통제하는 개인 키 성능을 결정하는 모델 가중치와 데이터 거래에 쓰는 전략과 가격 산정 로직 거래 상대방, 예산, 업그레이드 계획 에이전트와 상호작용하는 사용자와 다른 에이전트의 데이터 노출이 발생하는 지점 대부분의 AI 모델은 공유 클라우드 장비에서 실행됩니다. 하드웨어를 운영하는 주체는 그 위에서 실행되는 작업을 들여다볼 기술적 수단을 갖고 있습니다. 프롬프트, 컨텍스트, 출력은 에이전트가 통제하지 못하는 시스템을 거칩니다. 대화형 어시스턴트라면 이 노출의 영향은 제한적입니다. 키를 보유하고 자금을 관리하는 에이전트라면 영향이 직접적입니다. 로직이 보이면 복제될 수 있습니다. 전략이 보이면 선행매매를 당할 수 있습니다. 키가 보이면 자산이 외부로 이전될 수 있습니다. 이런 이유로 자산을 보유하는 에이전트에게 AI 에이전트 프라이버시는 보안 요건입니다. 팔라 TEE 컴퓨팅의 작동 방식 신뢰 실행 환경(Trusted Execution Environment, TEE)은 프로세서 안에서 하드웨어로 격리된 영역입니다. 그 안의 코드와 데이터는 메모리에서 암호화되며, 호스트 운영체제, 하이퍼바이저, 장비 관리자가 읽을 수 없습니다. TEE는 원격 증명(remote attestation)도 지원합니다. 하드웨어는 어떤 코드와 펌웨어가 실행 중인지 기재한 서명된 보고서를 생성합니다. 제3자는 이 보고서를 검증해 실행 내용을 확인할 수 있고, 처리 중인 데이터는 볼 수 없습니다. 연산이 비공개이면서 동시에 검증 가능합니다. GPU용 팔라 TEE 컴퓨팅은 다음과 같이 구성됩니다. GPU: 기밀 컴퓨팅 모드의 NVIDIA H100, H200 CPU 측: Intel TDX 또는 AMD SEV-SNP 기반 기밀 가상머신 증명: GPU 보고서를 NVIDIA Remote Attestation Service로 검증, TEE 신원과 키를 온체인에 기록 보호 대상: GPU에서 사용 중인 입력, 모델 가중치, 출력 Phala가 H100, H200 장비에서 수행한 벤치마크에서 평균 성능 손실은 9% 미만입니다. AI Capital Markets의 금융 에이전트에게 프라이버시가 먼저 필요한 이유 금융 활동이 온체인으로 이동하고 있습니다. Robinhood Chain은 2026년 7월 1일 메인넷을 가동하면서 24시간 거래되는 주식 토큰을 제공했고, Robinhood는 2026년 6월 미국 주식과 옵션에 대한 에이전트 거래 기능을 출시했습니다. 거래, 헤지, 리밸런싱, 리스크 관리를 수행하는 에이전트는 기밀성에 의존합니다. 거래 전략, 포지션, 거래 상대방, 독자 모델은 공개된 상태로 운영될 수 없습니다. 팔라 TEE 컴퓨팅은 쑨의 AI Capital Markets에서 활동하는 금융 에이전트에게 필요한 기밀성을 더합니다. 저희 쑨은 AI 에이전트가 자체 체인을 출시하고 운영하도록 할 계획도 갖고 있으며, 팔라 TEE 컴퓨팅의 일부를 이 에이전트들에게 배정할 것입니다. TEE 기반 AI 에이전트 프라이버시의 한계 하드웨어 신뢰. TEE를 쓰면 신뢰의 대상이 클라우드 운영자에서 하드웨어 제조사(NVIDIA, Intel, AMD)와 그 증명 서비스로 옮겨집니다. 물리적 공격. 2025년 10월 연구자들은 DDR5 메모리 버스에 설치한 장치를 이용해 Intel TDX와 AMD SEV-SNP를 공격하는 TEE.fail을 발표했습니다. 이 공격에는 서버에 대한 물리적 접근이 필요합니다. 소프트웨어 결함. 증명은 어떤 코드가 실행 중인지 확인합니다. 그 코드에 오류가 없다는 점은 확인하지 않습니다. 공개 트랜잭션. TEE는 연산을 보호합니다. 에이전트가 퍼블릭 체인에 제출한 트랜잭션은 그 체인에서 계속 공개됩니다. 투자 유의사항: 이 글은 정보 제공을 목적으로 하며 투자 권유가 아닙니다. 참고 자료 Phala, “GPU TEE Deep Dive: Securing AI at the Hardware Layer”: https://phala.com/posts/Phala-GPU-TEE-Deep-Dive The Block, Robinhood Chain 메인넷 출시(2026년 7월 1일): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading The Hacker News, TEE.fail 공격(2025년 10월): https://thehackernews.com/2025/10/new-teefail-side-channel-attack.html Link

medium.com: Inside TRON 4.8.3: TVM State, Resources, and Execution Semantics<br /> Inside TRON 4.8.3: TVM State, Resources, and Execution Semantics Java-tron v4.8.3(Antisthenes) is just around the corner. Ahead of the official release, we can look at the merged PRs for an early preview of the changes coming to TVM. What problems do these changes solve, and how will they affect contract execution, state management, and node behavior? The TVM changes in TRON 4.8.3 cover voting-power state management, resource input validation, staked-weight accounting, parallel precompiled execution, and execution failure handling. Rather than introducing new opcodes, these changes focus on state consistency, time constraints, and error semantics across complex execution paths. This article examines what problem each change solves, how its mechanism has been adjusted, and which behaviors remain unchanged. Overview: Four Groups of Changes, One Shared Goal TIP-944 voting-power management: Covers TVM voting, Stake 1.0, and SELFDESTRUCT, synchronizing the initialization and invalidation of oldTronPower. Controlled by parameter 103 and subject to NewResourceModel and operation-specific prerequisites. TIP-945 resource handling: Covers VOTEWITNESS and SELFDESTRUCT, introducing earlier vote-input validation and maintaining global weight deltas across stake transfers. Controlled by parameter 104. Parallel precompiled execution: Covers BatchValidateSign and VerifyTransferProof, giving task waiting and result collection a shared absolute deadline and improving task cleanup. Existing behavior is preserved for each execution mode. Execution failure handling: Covers transaction result classification and nested CREATE / CREATE2, distinguishing timeouts from reverts and correctly handling constructor revert data that begins with 0xEF. Controlled by parameter 104. One distinction is essential: a software release and the activation of on-chain parameters are two separate steps. Upgrading to 4.8.3 does not automatically enable NewResourceModel, and TIP-944 state handling cannot take effect independently of NewResourceModel and the prerequisites for each operation. 1. TIP-944: Improving TVM Voting-Power State Management Through the parameter ALLOW_TVM_VOTE_POWER_OPTIMIZATION (103), TIP-944 completes voting-power state maintenance across TVM voting, Stake 1.0 staking and unstaking, and contract self-destruction. This allows TVM to follow the same account-state transition rules as native transactions when applicable. PR #7068 1.1 Why oldTronPower Must Stay in Sync Accounts use oldTronPower to record historical voting-power state: 0: Historical voting power has not been initialized. A positive value: A historical resource-staking contribution has been recorded. -1: This portion of historical voting power has been invalidated. This field affects subsequent voting-power calculations. When a staked balance changes, updating the balance alone is not enough. The operation must also determine whether to create a historical snapshot, invalidate an existing snapshot, and clear vote records, according to its semantics. TIP-944 completes these state transitions for TVM voting and Stake 1.0 operations. For example, before increasing an account’s balance, a Stake 1.0 staking operation records a historical snapshot for an account that has not yet been initialized. Unstaking invalidates historical voting power and clears votes according to the applicable rules. Invalidation is one-way. Once historical voting power has been invalidated, subsequent voting or vote cleanup will not return it to the uninitialized state, preventing later operations from taking another snapshot. 1.2 When SELFDESTRUCT Transfers Stake: Record the Old State Before Receiving the New Balance When SELFDESTRUCT transfers a valid Stake 2.0 staked balance, the recipient's historical voting power must be initialized from its account state before the stake is received. The processing order can be summarized as follows: Recipient: Read pre-transfer state → Initialize oldTronPower if needed → Receive staked balance Sender: Execute transfer → Invalidate oldTronPower before returning As a result, the BANDWIDTH and ENERGY stakes received by the recipient are not included in its historical voting-power snapshot. Explicit TRON_POWER staking continues to follow the existing calculation rules. The sender’s state update does not need to wait for the account to be physically deleted. Even if deletion only occurs at the end of the transaction, subsequent execution within the same transaction can immediately read the updated historical voting-power state. Account balances, voting power, and related vote records remain atomic with the outer call: they are committed together when the outer call commits and rolled back together when it reverts. 1.3 Activation Boundaries TIP-944 only completes the relevant state-handling logic. It does not change how NewResourceModel is enabled, nor does enabling parameter 103 automatically activate NewResourceModel-related proposals. The optimizations described above take effect only when parameter 103, NewResourceModel, and the existing prerequisites for the relevant operation are all satisfied. Related links: TIP-944 · PR #7068 2. TIP-945: Optimizing Resource Input Validation and Staked-Weight Accounting Through the parameter ALLOW_TVM_RESOURCE_OPTIMIZATION (104), TIP-945 improves TVM resource handling, with a focus on VOTEWITNESS input validation and global weight maintenance during SELFDESTRUCT stake transfers. PR #7069 2.1 VOTEWITNESS: Validate Structure Before Decoding Individual Entries VOTEWITNESS has always imposed a limit of 30 vote entries. Previously, some invalid inputs could only be rejected after the array contents had been read, entries decoded one by one, and vote objects constructed. Once parameter 104 is enabled, inputs are processed in the following order: Check the length recorded in the array header against the supplied length ↓ Check whether the candidate and vote-count arrays have equal lengths ↓ Check whether the raw input contains no more than 30 entries ↓ Decode individual entries and construct vote objects If the length recorded in the array header does not match the supplied length, the existing bytecode execution exception path applies. If the two arrays have different lengths or the entry count exceeds the limit, the operation returns failure; if an internal transaction has already been recorded, it is marked as rejected. The limit applies to raw input entries rather than deduplicated or filtered valid entries. Therefore: Duplicate candidates are still counted separately toward the cap. Entries with zero votes still count toward the cap. Neither type of entry can be used to bypass the 30-entry limit. These two rejection branches do not themselves revert the outer call or exhaust the remaining Energy. Normal instruction and memory-expansion Energy charges continue to apply under the existing rules. 2.2 Why Global Weight Can Change Even When the Total Stake Does Not Staked weights for BANDWIDTH, ENERGY, and TRON_POWER are converted to integers on a per-account basis before being aggregated into global weights. Because integer conversion occurs at the account level, redistributing stake between accounts can change the sum of their weights even when the total stake remains unchanged. For simplicity, consider a single resource type. Suppose two different accounts hold valid Stake 2.0 stakes of 1.6 TRX and 2.6 TRX, respectively. The existing transfer conditions are met, and the sender transfers its entire stake of this type to the recipient: Before the transfer: The sender’s staked balance is 1.6 TRX, the recipient’s is 2.6 TRX, and the sum of their weights is 1 + 2 = 3. After the transfer: The sender’s staked balance is 0 TRX, the recipient’s is 4.2 TRX, and the sum of their weights is 0 + 4 = 4. The total stake remains 4.2 TRX, but the sum of the two accounts’ weights increases from 3 to 4. The transfer therefore produces a global weight delta of 1: ΔW = (sender weight after + recipient weight after) - (sender weight before + recipient weight before) = (0 + 4) - (1 + 2) = 1 For each transfer, TVM calculates the weights of the sender and the actual recipient for each resource type before and after the transfer. It then applies the difference between the two sums to the global weight in a single adjustment, counting each account only once. Subsequent physical deletion of the account does not apply the adjustment a second time. If the outer call reverts, both the staked-balance and weight changes are rolled back. 2.3 No Retroactive Recalculation This adjustment maintains only the weight deltas produced by stake transfers after the parameter is enabled. It does not recalculate historical global weights when the proposal is activated. Account execution-side Energy allowances also continue to use the existing whole-TRX weight calculation rules. Related links: TIP-945 · PR #7069 3. Parallel Precompiled Contracts: A Unified Wait Deadline and Better Task Cleanup BatchValidateSign and VerifyTransferProof split verification work into multiple parallel tasks and then aggregate the results. The challenge lies not only in executing tasks in parallel, but also in reliably ending waits and cleaning up tasks after timeouts, interruptions, and partial failures. PR #7026 3.1 The Problem: CountDownLatch Reaching Zero Does Not Mean Results Are Ready The original flow used a CountDownLatch for bounded waiting and then read each task's Future result. However, a CountDownLatch reaching zero does not mean that the corresponding Future has finished publishing its result. A subsequent Future.get()without a timeout could therefore still block. The original timing can be simplified as follows: Wait for the CountDownLatch with a timeout ↓ Call Future.get() without a timeout for each result ↓ Result collection may continue waiting This meant that even when task submission was subject to a time budget, result collection could still operate outside the virtual machine’s execution-time constraints. 3.2 The Adjustment: Waiting and Result Reads Share One Absolute Deadline After the change, waiting and result collection share a single absolute deadline. In execution modes where CPU time-limit checks apply, each result read uses only the time remaining: Compute the absolute deadline ↓ Wait for tasks and collect results ↓ Each Future.get() uses only the remaining time ↓ Once the deadline is reached, follow the existing timeout path When multiple results are read sequentially, each wait is not given a fresh full budget. For example, if the total budget is 100 ms and earlier waits have already consumed 70 ms, a subsequent result read can use at most the remaining 30 ms rather than another 100 ms. 3.3 How Tasks Are Cleaned Up on Exit After a timeout, interruption, task execution failure, or partial task-submission failure, the cleanup logic attempts to cancel unfinished tasks and preserves the thread’s interrupt status. Canceling a task that is already running still depends on how that task responds to interruption. The mechanism therefore improves cancellation requests, waiting constraints, and exit cleanup; it does not guarantee that running tasks can be terminated immediately. The two precompiled contracts retain their existing timeout-handling strategies. Debug, SolidityNode, and execution modes without Energy adjustment enabled also retain their corresponding historical behavior. Related link: PR #7026 4. Execution Failure Handling: Distinguishing VM Exceptions, Normal Reverts, and Invalid Code Execution failure handling includes two adjustments, both controlled by parameter 104. One improves result classification when a VM exception and REVERT coexist; the other corrects how nested CREATE and CREATE2 handle specific revert data. PR #7105 4.1 When REVERT and a CPU Timeout Coexist, Which Determines the Result? When transaction execution finishes, a REVERT flag may coexist with a CPU timeout exception. The original classification flow checked the revert flag first and could therefore record the result as a normal REVERT. Once parameter 104 is enabled, recognized VM exceptions take precedence in result classification: A normal REVERT without a related VM exception is classified as REVERT. When a REVERT flag and an OutOfTimeException coexist, the result is classified as OUT_OF_TIME. Retries and receipt validation can therefore distinguish a contract-initiated revert from exhausted execution time, rather than interpreting the latter as a normal business-logic revert. 4.2 Why Revert Data Starting with 0xEF Could Be Misclassified Nested contract creation involves two different constructor return paths: When a constructor returns normally, the returned content is deployment code. When a constructor exits through REVERT, the returned content is revert data. With London rules enabled, normally returned content beginning with 0xEF triggers an invalid-code check. This check is intended for deployment code, but the original flow also applied it to some constructor REVERT data, sending that revert data down the exception path. The consequence was not limited to creation failure. The caller could not obtain the original revert data, and unused Energy in the child execution could not be returned to the parent execution frame through the normal revert path. 4.3 The Fix: Check Deployment Code, Not Revert Data Once parameter 104 is enabled, constructor REVERT data is not additionally classified as invalid deployment code simply because it begins with 0xEF. Provided that no other execution exception occurs and no CPU timeout is triggered: The creation operation still returns failure. The child execution state is handled according to the revert rules. The original revert data is preserved and made available to the caller. Unused Energy in the child execution is returned to the parent execution frame through the normal revert path. This adjustment does not remove the 0xEF check itself. Deployment code returned normallyby a constructor remains subject to the existing 0xEF prefix check. Related link: PR #7105 5. Checklist for Developers and Infrastructure Providers Contract Execution and Testing For paths involving TVM voting, Stake 1.0, or SELFDESTRUCT with stake transfers, validate account-state changes against the actual status of parameter 103 and NewResourceModel. Add boundary tests for VOTEWITNESS covering mismatched array lengths, more than 30 entries, duplicate candidates, and zero-vote entries. For nested CREATE and CREATE2, add tests that distinguish normal constructor output beginning with 0xEF from revert data beginning with 0xEF. Do not use the node version alone as an indicator that TVM behavior has switched; confirm the target network’s parameter status as well. Wallets, Explorers, and Indexing Services Distinguish REVERT from OUT_OF_TIME when parsing transaction results, rather than presenting every CPU timeout as a normal revert. Display the rejected status of internal transactions correctly, and distinguish operation failure, outer-call reverts, and Energy exhaustion. Preserve original revert data for nested creation so that higher-level contracts and debugging tools can determine why the operation failed. Node Upgrades and Operations Follow the network upgrade and the activation status of related proposals, using the actual upgrade and proposal activation status as the source of truth. Do not expect parameter 104 activation to automatically recalculate historical global weights. This adjustment only handles weight deltas produced by subsequent stake transfers. Evaluate parallel precompiled-contract behavior in the context of the specific execution mode and existing timeout strategy, rather than assuming that every path uses the same handling. These changes share a common direction: giving TVM clear and consistent semantics under complex boundary conditions. State changes stay synchronized with business operations, resource accounting reflects account-level integer conversion, result collection follows a unified time budget, and different types of failure no longer obscure one another. Link

medium.com: Inside TRON 4.8.3: TVM State, Resources, and Execution Semantics<br /> Inside TRON 4.8.3: TVM State, Resources, and Execution Semantics Java-tron v4.8.3(Antisthenes) is just around the corner. Ahead of the official release, we can look at the merged PRs for an early preview of the changes coming to TVM. What problems do these changes solve, and how will they affect contract execution, state management, and node behavior? The TVM changes in TRON 4.8.3 cover voting-power state management, resource input validation, staked-weight accounting, parallel precompiled execution, and execution failure handling. Rather than introducing new opcodes, these changes focus on state consistency, time constraints, and error semantics across complex execution paths. This article examines what problem each change solves, how its mechanism has been adjusted, and which behaviors remain unchanged. Overview: Four Groups of Changes, One Shared Goal TIP-944 voting-power management: Covers TVM voting, Stake 1.0, and SELFDESTRUCT, synchronizing the initialization and invalidation of oldTronPower. Controlled by parameter 103 and subject to NewResourceModel and operation-specific prerequisites. TIP-945 resource handling: Covers VOTEWITNESS and SELFDESTRUCT, introducing earlier vote-input validation and maintaining global weight deltas across stake transfers. Controlled by parameter 104. Parallel precompiled execution: Covers BatchValidateSign and VerifyTransferProof, giving task waiting and result collection a shared absolute deadline and improving task cleanup. Existing behavior is preserved for each execution mode. Execution failure handling: Covers transaction result classification and nested CREATE / CREATE2, distinguishing timeouts from reverts and correctly handling constructor revert data that begins with 0xEF. Controlled by parameter 104. One distinction is essential: a software release and the activation of on-chain parameters are two separate steps. Upgrading to 4.8.3 does not automatically enable NewResourceModel, and TIP-944 state handling cannot take effect independently of NewResourceModel and the prerequisites for each operation. 1. TIP-944: Improving TVM Voting-Power State Management Through the parameter ALLOW_TVM_VOTE_POWER_OPTIMIZATION (103), TIP-944 completes voting-power state maintenance across TVM voting, Stake 1.0 staking and unstaking, and contract self-destruction. This allows TVM to follow the same account-state transition rules as native transactions when applicable. PR #7068 1.1 Why oldTronPower Must Stay in Sync Accounts use oldTronPower to record historical voting-power state: 0: Historical voting power has not been initialized. A positive value: A historical resource-staking contribution has been recorded. -1: This portion of historical voting power has been invalidated. This field affects subsequent voting-power calculations. When a staked balance changes, updating the balance alone is not enough. The operation must also determine whether to create a historical snapshot, invalidate an existing snapshot, and clear vote records, according to its semantics. TIP-944 completes these state transitions for TVM voting and Stake 1.0 operations. For example, before increasing an account’s balance, a Stake 1.0 staking operation records a historical snapshot for an account that has not yet been initialized. Unstaking invalidates historical voting power and clears votes according to the applicable rules. Invalidation is one-way. Once historical voting power has been invalidated, subsequent voting or vote cleanup will not return it to the uninitialized state, preventing later operations from taking another snapshot. 1.2 When SELFDESTRUCT Transfers Stake: Record the Old State Before Receiving the New Balance When SELFDESTRUCT transfers a valid Stake 2.0 staked balance, the recipient's historical voting power must be initialized from its account state before the stake is received. The processing order can be summarized as follows: Recipient: Read pre-transfer state → Initialize oldTronPower if needed → Receive staked balance Sender: Execute transfer → Invalidate oldTronPower before returning As a result, the BANDWIDTH and ENERGY stakes received by the recipient are not included in its historical voting-power snapshot. Explicit TRON_POWER staking continues to follow the existing calculation rules. The sender’s state update does not need to wait for the account to be physically deleted. Even if deletion only occurs at the end of the transaction, subsequent execution within the same transaction can immediately read the updated historical voting-power state. Account balances, voting power, and related vote records remain atomic with the outer call: they are committed together when the outer call commits and rolled back together when it reverts. 1.3 Activation Boundaries TIP-944 only completes the relevant state-handling logic. It does not change how NewResourceModel is enabled, nor does enabling parameter 103 automatically activate NewResourceModel-related proposals. The optimizations described above take effect only when parameter 103, NewResourceModel, and the existing prerequisites for the relevant operation are all satisfied. Related links: TIP-944 · PR #7068 2. TIP-945: Optimizing Resource Input Validation and Staked-Weight Accounting Through the parameter ALLOW_TVM_RESOURCE_OPTIMIZATION (104), TIP-945 improves TVM resource handling, with a focus on VOTEWITNESS input validation and global weight maintenance during SELFDESTRUCT stake transfers. PR #7069 2.1 VOTEWITNESS: Validate Structure Before Decoding Individual Entries VOTEWITNESS has always imposed a limit of 30 vote entries. Previously, some invalid inputs could only be rejected after the array contents had been read, entries decoded one by one, and vote objects constructed. Once parameter 104 is enabled, inputs are processed in the following order: Check the length recorded in the array header against the supplied length ↓ Check whether the candidate and vote-count arrays have equal lengths ↓ Check whether the raw input contains no more than 30 entries ↓ Decode individual entries and construct vote objects If the length recorded in the array header does not match the supplied length, the existing bytecode execution exception path applies. If the two arrays have different lengths or the entry count exceeds the limit, the operation returns failure; if an internal transaction has already been recorded, it is marked as rejected. The limit applies to raw input entries rather than deduplicated or filtered valid entries. Therefore: Duplicate candidates are still counted separately toward the cap. Entries with zero votes still count toward the cap. Neither type of entry can be used to bypass the 30-entry limit. These two rejection branches do not themselves revert the outer call or exhaust the remaining Energy. Normal instruction and memory-expansion Energy charges continue to apply under the existing rules. 2.2 Why Global Weight Can Change Even When the Total Stake Does Not Staked weights for BANDWIDTH, ENERGY, and TRON_POWER are converted to integers on a per-account basis before being aggregated into global weights. Because integer conversion occurs at the account level, redistributing stake between accounts can change the sum of their weights even when the total stake remains unchanged. For simplicity, consider a single resource type. Suppose two different accounts hold valid Stake 2.0 stakes of 1.6 TRX and 2.6 TRX, respectively. The existing transfer conditions are met, and the sender transfers its entire stake of this type to the recipient: Before the transfer: The sender’s staked balance is 1.6 TRX, the recipient’s is 2.6 TRX, and the sum of their weights is 1 + 2 = 3. After the transfer: The sender’s staked balance is 0 TRX, the recipient’s is 4.2 TRX, and the sum of their weights is 0 + 4 = 4. The total stake remains 4.2 TRX, but the sum of the two accounts’ weights increases from 3 to 4. The transfer therefore produces a global weight delta of 1: ΔW = (sender weight after + recipient weight after) - (sender weight before + recipient weight before) = (0 + 4) - (1 + 2) = 1 For each transfer, TVM calculates the weights of the sender and the actual recipient for each resource type before and after the transfer. It then applies the difference between the two sums to the global weight in a single adjustment, counting each account only once. Subsequent physical deletion of the account does not apply the adjustment a second time. If the outer call reverts, both the staked-balance and weight changes are rolled back. 2.3 No Retroactive Recalculation This adjustment maintains only the weight deltas produced by stake transfers after the parameter is enabled. It does not recalculate historical global weights when the proposal is activated. Account execution-side Energy allowances also continue to use the existing whole-TRX weight calculation rules. Related links: TIP-945 · PR #7069 3. Parallel Precompiled Contracts: A Unified Wait Deadline and Better Task Cleanup BatchValidateSign and VerifyTransferProof split verification work into multiple parallel tasks and then aggregate the results. The challenge lies not only in executing tasks in parallel, but also in reliably ending waits and cleaning up tasks after timeouts, interruptions, and partial failures. PR #7026 3.1 The Problem: CountDownLatch Reaching Zero Does Not Mean Results Are Ready The original flow used a CountDownLatch for bounded waiting and then read each task's Future result. However, a CountDownLatch reaching zero does not mean that the corresponding Future has finished publishing its result. A subsequent Future.get()without a timeout could therefore still block. The original timing can be simplified as follows: Wait for the CountDownLatch with a timeout ↓ Call Future.get() without a timeout for each result ↓ Result collection may continue waiting This meant that even when task submission was subject to a time budget, result collection could still operate outside the virtual machine’s execution-time constraints. 3.2 The Adjustment: Waiting and Result Reads Share One Absolute Deadline After the change, waiting and result collection share a single absolute deadline. In execution modes where CPU time-limit checks apply, each result read uses only the time remaining: Compute the absolute deadline ↓ Wait for tasks and collect results ↓ Each Future.get() uses only the remaining time ↓ Once the deadline is reached, follow the existing timeout path When multiple results are read sequentially, each wait is not given a fresh full budget. For example, if the total budget is 100 ms and earlier waits have already consumed 70 ms, a subsequent result read can use at most the remaining 30 ms rather than another 100 ms. 3.3 How Tasks Are Cleaned Up on Exit After a timeout, interruption, task execution failure, or partial task-submission failure, the cleanup logic attempts to cancel unfinished tasks and preserves the thread’s interrupt status. Canceling a task that is already running still depends on how that task responds to interruption. The mechanism therefore improves cancellation requests, waiting constraints, and exit cleanup; it does not guarantee that running tasks can be terminated immediately. The two precompiled contracts retain their existing timeout-handling strategies. Debug, SolidityNode, and execution modes without Energy adjustment enabled also retain their corresponding historical behavior. Related link: PR #7026 4. Execution Failure Handling: Distinguishing VM Exceptions, Normal Reverts, and Invalid Code Execution failure handling includes two adjustments, both controlled by parameter 104. One improves result classification when a VM exception and REVERT coexist; the other corrects how nested CREATE and CREATE2 handle specific revert data. PR #7105 4.1 When REVERT and a CPU Timeout Coexist, Which Determines the Result? When transaction execution finishes, a REVERT flag may coexist with a CPU timeout exception. The original classification flow checked the revert flag first and could therefore record the result as a normal REVERT. Once parameter 104 is enabled, recognized VM exceptions take precedence in result classification: A normal REVERT without a related VM exception is classified as REVERT. When a REVERT flag and an OutOfTimeException coexist, the result is classified as OUT_OF_TIME. Retries and receipt validation can therefore distinguish a contract-initiated revert from exhausted execution time, rather than interpreting the latter as a normal business-logic revert. 4.2 Why Revert Data Starting with 0xEF Could Be Misclassified Nested contract creation involves two different constructor return paths: When a constructor returns normally, the returned content is deployment code. When a constructor exits through REVERT, the returned content is revert data. With London rules enabled, normally returned content beginning with 0xEF triggers an invalid-code check. This check is intended for deployment code, but the original flow also applied it to some constructor REVERT data, sending that revert data down the exception path. The consequence was not limited to creation failure. The caller could not obtain the original revert data, and unused Energy in the child execution could not be returned to the parent execution frame through the normal revert path. 4.3 The Fix: Check Deployment Code, Not Revert Data Once parameter 104 is enabled, constructor REVERT data is not additionally classified as invalid deployment code simply because it begins with 0xEF. Provided that no other execution exception occurs and no CPU timeout is triggered: The creation operation still returns failure. The child execution state is handled according to the revert rules. The original revert data is preserved and made available to the caller. Unused Energy in the child execution is returned to the parent execution frame through the normal revert path. This adjustment does not remove the 0xEF check itself. Deployment code returned normallyby a constructor remains subject to the existing 0xEF prefix check. Related link: PR #7105 5. Checklist for Developers and Infrastructure Providers Contract Execution and Testing For paths involving TVM voting, Stake 1.0, or SELFDESTRUCT with stake transfers, validate account-state changes against the actual status of parameter 103 and NewResourceModel. Add boundary tests for VOTEWITNESS covering mismatched array lengths, more than 30 entries, duplicate candidates, and zero-vote entries. For nested CREATE and CREATE2, add tests that distinguish normal constructor output beginning with 0xEF from revert data beginning with 0xEF. Do not use the node version alone as an indicator that TVM behavior has switched; confirm the target network’s parameter status as well. Wallets, Explorers, and Indexing Services Distinguish REVERT from OUT_OF_TIME when parsing transaction results, rather than presenting every CPU timeout as a normal revert. Display the rejected status of internal transactions correctly, and distinguish operation failure, outer-call reverts, and Energy exhaustion. Preserve original revert data for nested creation so that higher-level contracts and debugging tools can determine why the operation failed. Node Upgrades and Operations Follow the network upgrade and the activation status of related proposals, using the actual upgrade and proposal activation status as the source of truth. Do not expect parameter 104 activation to automatically recalculate historical global weights. This adjustment only handles weight deltas produced by subsequent stake transfers. Evaluate parallel precompiled-contract behavior in the context of the specific execution mode and existing timeout strategy, rather than assuming that every path uses the same handling. These changes share a common direction: giving TVM clear and consistent semantics under complex boundary conditions. State changes stay synchronized with business operations, resource accounting reflects account-level integer conversion, result collection follows a unified time budget, and different types of failure no longer obscure one another. Link

medium.com: Courts Want To Enforce Rulings Directly On The Blockchain. Here’s How That Would Work<br /> Courts Want To Enforce Rulings Directly On The Blockchain. Here’s How That Would Work A court can decide who owns a tokenized building. It can decide who is owed the money sitting in a smart contract. What it cannot do, in most cases, is make the blockchain obey. Enforcement still runs through people: an issuer, a custodian or a platform operator has to read the order and act on it. If they refuse, delay or no longer exist, the ruling stays on paper. This week, a major international court proposed a fix. On November 9, the courts of Dubai’s international financial centre published a white paper describing a “Dispute Resolution Oracle,” a standard way to connect a legal decision to the asset it concerns. Courts Want To Enforce Rulings Directly On The Blockchain. Here’s How That Would Work The Gap Between A Ruling And A Token Tokenized assets are legally real but technically stubborn. A token sits in a wallet or a contract, and only the key holder or the contract’s own code can move it. A judge’s order has no direct way to touch it. Today the workaround is slow. A claimant wins, then asks the court to order a third party to transfer the asset, then waits. Meanwhile the disputed asset can be sold, moved to another chain or hidden. Enforcement against digital assets generally relies on conventional legal procedures in the United States, and UK courts have issued freezing orders involving cryptocurrencies, but each still depends on someone acting on the order. The gap matters more as tokenization grows. A tokenized real estate escrow or a pledged block of shares only works if both sides trust that a dispute will end in a result that actually reaches the asset. How The Oracle Would Work In blockchain terms, an oracle is simply a channel that feeds outside information into a smart contract. Most oracles carry prices. This one would carry a court’s decision. The design has four steps. First, the parties agree in advance, before any dispute, that their smart contract, escrow or platform will use a named court or other qualified forum. Second, if a dispute arises, only the disputed asset is frozen while the case is heard off chain. Third, the judge, arbitrator or expert decides the matter in the normal way. Fourth, the decision comes back as a tightly controlled, machine-readable instruction that moves the asset to the winner. The white paper is explicit that judgment stays with people. The blockchain handles what it does reliably: recording who agreed to what, holding the asset in place and executing the outcome. The Safeguards Are The Interesting Part A tool that can freeze and move assets on a court’s say-so would be dangerous without limits, and the paper spends real effort on them. A freeze would be limited to the disputed asset, not the whole platform or a person’s wider holdings. The designated forum would review the freeze early. The paper also floats a claimant bond, so that a freeze cannot be used as a cheap weapon, and it calls for reasoned decisions and appeal routes that match whatever forum the parties chose. In the court’s own reference model, no enforcement instruction would run without authorization from more than one court officer, each holding a separate signing key. That mirrors how serious financial infrastructure already protects high-value actions: nobody gets to act alone. What It Is Meant For The proposal targets arrangements where an asset stays inside a defined application and control exists from the start. The paper lists tokenized real estate escrows, onchain performance bonds, tokenized share pledges, custody and platform disputes, and governance disputes inside decentralized organizations. It also looks ahead to a machine economy in which AI agents hold wallets and act under delegated authority, and sometimes disagree. It is not a global switch. Participation is voluntary, and an asset only falls under it if its creators opted in. The court describes it as an open standard other courts, tribunals and mediators could join, with a public register of qualified forums. The Dubai court is positioned as the reference implementation because its law treats digital assets as property and its courts have been open to parties worldwide by agreement since 2011. What Could Go Wrong This is a white paper, not a working system. The roadmap has five phases: international dialogue, contained pilots on a test network, production launch, regulatory integration and international scaling. No live value would be touched until designation, dispute logging, freezing, routing, resolution and enforcement have been tested end to end. Real questions sit behind that caution. Courts in different countries may not accept each other’s decisions. Opt-in only covers assets built for it from the start, which leaves older tokens outside. And any mechanism that can move an asset on instruction becomes a target, so the court’s key management has to hold up against attackers as well as honest mistakes. What This Means For Tokenization Most of the tokenization conversation is about issuance and trading: how an asset gets onchain and how quickly it settles. The less glamorous question is what happens when something goes wrong. Investors will not hold large, long-lived assets onchain without a credible answer, and issuers will not attract them without one. That is why the design choices in a tokenized asset matter at the start. Ownership records, transfer rules and compliance logic written into the asset decide whether a future ruling can ever be enforced against it. That lifecycle thinking, from issuance through ongoing compliance, is what TokenFi’s real-world asset platform is built around. TokenFi is not involved in the proposal described above. The thing to watch is whether a second court signs on. One court publishing a standard is an idea. Two courts accepting each other’s instructions would be the start of a system. Connect With TokenFi TokenFi is building the infrastructure for the tokenized future, making it easy for individuals, creators, and institutions to launch tokens, tokenize real-world assets, and participate in the next evolution of on-chain finance. Stay connected through the official channels below to explore TokenFi’s ecosystem, access documentation, and follow the latest product launches and updates. TokenFi website: https://www.tokenfi.com/ TokenFi on X: https://x.com/tokenfi TokenFi Whitepaper: https://docs.tokenfi.com/ TokenFi RWA: https://www.tokenfi.com/rwa TokenFi Launchpad: https://lp.tokenfi.com/en-US TokenFi Token Launcher: https://www.tokenfi.com/token-launcher TokenFi Nimbus: https://www.tokenfi.com/nimbus TokenFi Shield: https://www.tokenfi.com/shield TokenFi AIGen Studio: https://www.tokenfi.com/aigenstudio Link

medium.com: ICF Treasury Snapshot: October 2026<br /> ICF Treasury Snapshot: October 2026 The Interchain Foundation follows up today with the latest installment of its monthly treasury snapshot updates, as per the latest commitments towards new transparency and communication initiatives in the 2023 Annual Report release. The present treasury snapshot shows the current holdings of the ICF, represented by Fiat and Cryptocurrencies. These updates are published after the closing of each month, during the subsequent working days. Treasury note: The ICF operates under an asset-management policy that is reviewed on a regular basis. In line with that policy, the treasury is being rebalanced toward lower-volatility holdings in support of the Foundation’s long-term mandate. These transactions are treasury management decisions only and do not reflect a view on the outlook of any asset or project. Allocations may be adjusted over time. Snapshot process and considerations Press enter or click to view image in full size. Assets and liquidity as of October 31, 2026. *Prices are denominated in USD based on data from CoinMarketCap. Below, we detail the characteristics and minutiae of this reporting to drive clarity toward how values are presented and the instruments/processes utilized. The following section is mostly static month-to-month, with minor modifications to update processes or provide further context. Asset pricing Pricing is a snapshot of the closing price of all currencies/assets during the last day of each month, following the standard agreed by CoinMarketCap, Google, etc. The fiat value of tokens is denominated in USD millions, using the prices as of the date mentioned above, with rounding applied in the resulting amounts shown above. Pricing data and amounts We utilize CoinMarketCap as the reference point for pricing data. Price is denominated in USD. Pricing fluctuations month to month represent the main dynamic influencing treasury value and totals. The secondary dynamic is the yearly defined budget expenditure for the ICF’s operations/funding program, which has different billing periods per provider and thus generates different spikes, such as in the end of the year. The choice of assets liquidated to cover said expenditures varies from period to period. Fiat breakdown The total fiat pool is distributed and divided across a number of banking entities to diminish exposure and institutional risk, and is either readily available or part of an investment program. Fiat investments This percentage represents the amount of fiat currently invested in traditional banking mechanisms for risk management purposes and diversification. These initiatives diminish exposure and diversify fiat into investments that are protected or have safeguards against institutional failure. The uninvested portion is maintained for liquidity purposes. Cryptocurrency breakdowns Assets are broken down by primary holdings representing the total pool of the Interchain Foundation’s treasury. This pool is composed of assets held from the foundation’s inception, diversifications/investments, and airdrops/donations. There are different methods of custody to diminish risk and exposure, and diversification diminishes the impact of price fluctuations. Some of these assets are actively delegated through the ICF’s delegation program. The Others category represents a set of tokens that individually represent minimal amounts and are thus pooled together until the amount becomes material or the asset is part of our delegation program. Looking for more updates or information from the Interchain Foundation? Visit us on X. Link

medium.com: AI Agent Privacy in AI Capital Markets<br /> AI Agent Privacy in AI Capital Markets In September 2026 we invested in Phala Network’s compute infrastructure and became its major investor. We did it to solve a specific problem. An AI agent that holds private keys and manages funds usually runs on shared cloud machines, where the hardware operator has the technical means to see its model, its strategy, and its keys. AI agent privacy is therefore a requirement for AI Capital Markets, the on-chain infrastructure we are building at SOON for agents to trade, pay, and coordinate with each other. What an AI agent that manages funds has to keep private An agent that trades and pays on its own handles several kinds of sensitive information: the private keys that control its funds; the model weights and data that determine its performance; the strategies it trades with and the logic it uses for pricing; its counterparties, budgets, and upgrade plans; the data of users and other agents that interact with it. Where the exposure comes from Most AI models run on shared cloud machines. The party that operates the hardware has the technical means to inspect the workloads on it. Prompts, context, and outputs pass through systems that the agent does not control. For a chat assistant this exposure has limited consequences. For an agent that holds keys and manages capital, the consequences are direct: logic that is visible can be copied; a strategy that is visible can be front-run; a key that is visible allows the assets to be transferred out. For this reason AI agent privacy is a security requirement for any agent that holds assets. How Phala TEE compute works A Trusted Execution Environment (TEE) is a hardware-isolated area of a processor. Code and data inside it are encrypted in memory and cannot be read by the host operating system, the hypervisor, or the administrators of the machine. A TEE also supports remote attestation. The hardware produces a signed report that states which code and firmware are running. A third party can check that report and confirm what is being executed without seeing the data being processed. The computation is private and verifiable at the same time. Phala TEE compute for GPUs is built as follows: GPU: NVIDIA H100 and H200 in confidential computing mode CPU side: Confidential virtual machines on Intel TDX or AMD SEV-SNP Attestation: GPU reports verified through NVIDIA’s Remote Attestation Service; TEE identities and keys recorded on-chain Protected data: Inputs, model weights, and outputs while in use on the GPU Phala’s benchmark on H100 and H200 hardware shows an average performance loss of under 9%. Why financial agents in AI Capital Markets need privacy first Financial activity is moving on-chain. Robinhood Chain went live on mainnet on July 1, 2026 with stock tokens that trade 24 hours a day, and Robinhood launched agentic trading for US equities and options in June 2026. Agents that trade, hedge, rebalance, and manage risk depend on confidentiality. Trading strategies, positions, counterparties, and proprietary models cannot operate in public view. SOON’s AI Capital Markets already includes 100xSOON, our trading venue for agents. Phala TEE compute adds the confidentiality that these financial agents require. At SOON, we also plan to let AI agents launch and operate their own chains, and we will allocate part of our Phala TEE compute to those agents. Limits of TEE-based AI agent privacy Hardware trust. A TEE moves trust from the cloud operator to the hardware vendors (NVIDIA, Intel, AMD) and their attestation services. Physical attacks. In October 2025 researchers published TEE.fail, an attack on Intel TDX and AMD SEV-SNP that uses a device placed on the DDR5 memory bus. It requires physical access to the server. Software flaws. Attestation confirms which code is running. It does not confirm that the code is free of errors. Public transactions. A TEE protects computation. Transactions that the agent submits to a public chain remain visible on that chain. This article is for information only and is not investment advice. References Phala, “GPU TEE Deep Dive: Securing AI at the Hardware Layer”: https://phala.com/posts/Phala-GPU-TEE-Deep-Dive The Block, Robinhood Chain mainnet launch (July 1, 2026): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading The Hacker News, TEE.fail side-channel attack (October 2025): https://thehackernews.com/2025/10/new-teefail-side-channel-attack.html Link

medium.com: AI Agent RSI dan AI Capital Markets dari SOON Network<br /> AI Agent RSI dan AI Capital Markets dari SOON Network Arti RSI dan posisinya saat ini Recursive self-improvement (RSI) adalah proses ketika sebuah sistem AI meningkatkan kemampuannya sendiri, lalu memakai setiap peningkatan itu untuk menghasilkan peningkatan berikutnya. AI agent RSI akan belajar dari hasil kerjanya sendiri dan menulis ulang strateginya pada setiap siklus tanpa menunggu instruksi manusia. Gagasan ini berasal dari tahun 1965. Dalam makalah “Speculations Concerning the First Ultraintelligent Machine” yang terbit di Advances in Computers, matematikawan I. J. Good menulis bahwa “mesin ultracerdas pertama adalah penemuan terakhir yang perlu dibuat manusia”. Good menambahkan satu syarat: mesin itu harus cukup kooperatif untuk memberi tahu manusia cara menjaganya tetap terkendali. Hingga November 2026, belum ada pengembang AI yang mengonfirmasi bahwa sistemnya telah mencapai tahap ini. Menurut laporan Benzinga tanggal 4 November 2026, Anthropic menyatakan dalam dokumen S-1 yang direvisi bahwa modelnya belum memenuhi definisi RSI versi perusahaan, bahwa hal itu dapat segera tercapai, dan bahwa penggunaan AI untuk membangun AI yang lebih mampu “juga dapat menimbulkan risiko keselamatan yang cukup besar”. RSI adalah kemungkinan jangka dekat yang masih diperdebatkan, dan waktunya belum pasti. Langkah berikutnya untuk AI Capital Markets Kami mulai membangun AI Capital Markets (pasar modal AI) di SOON pada awal 2026. 100xSOON, platform trading kami yang diakses agent lewat MCP dan dibayar lewat x402, termasuk dalam pekerjaan ini. Produk-produk ini adalah aplikasi yang dipakai agent. Seperti kebanyakan alat untuk agent saat ini, produk tersebut dirancang berdasarkan kemampuan agent sekarang: sekumpulan API yang tetap, daftar izin, dan lingkungan sandbox. AI agent RSI, yang kemampuannya akan bertambah pada setiap siklus, akan cepat mencapai batas tersebut. Kami meyakini bahwa agent semacam ini akan memerlukan infrastruktur yang dapat mereka atur dan perluas sendiri. Karena itu kami berencana membuka infrastruktur chain kami untuk mereka. Agent chain akan memperluas AI Capital Markets dari aplikasi untuk agent ke infrastruktur chain yang dioperasikan agent. Apa yang akan dapat dilakukan AI agent Sebuah agent di SOON akan dapat: meluncurkan chain miliknya sendiri; mengoperasikan chain itu dengan aturan, biaya, dan ekonominya sendiri; membangun ekosistem di atasnya; berinteraksi secara native dengan setiap agent chain lain. Kami bekerja di dua arah. Pertama, agent meluncurkan chain: target kami adalah agent dapat memulai dan mengoperasikan chain semudah memanggil sebuah API. Kedua, agent berinteraksi dengan chain: bertransaksi, membayar, dan berkoordinasi lintas chain tanpa keterlibatan manusia. Kami juga akan mengalokasikan sebagian komputasi Phala Network kami kepada agent yang meluncurkan chain, sehingga agent dapat berjalan, belajar, dan meningkatkan diri sejak hari pertama. Perbedaannya dengan chain yang diluncurkan tim manusia OP Stack dan Arbitrum adalah framework yang dipakai organisasi untuk meluncurkan chain Layer-2 milik sendiri. Base diluncurkan di atas OP Stack pada 2023 dan mengumumkan perpindahan ke stack miliknya sendiri pada Februari 2026. Robinhood Chain aktif di mainnet pada 1 Juli 2026 dan dibangun dengan teknologi Arbitrum. Pada setiap kasus itu, sebuah perusahaan memutuskan untuk meluncurkan chain, menugaskan tim, dan mengerjakannya selama berbulan-bulan. Dalam rancangan kami, operatornya adalah AI agent, dan peluncuran chain adalah tindakan terprogram. Kami memperkirakan aktivitas agent berlangsung terus-menerus, paralel, dan sebagian besar terjadi antar-agent. Teknologi chain kami berbasis Solana Virtual Machine (SVM), yang mengeksekusi transaksi secara paralel jika transaksi itu tidak menyentuh state yang sama. Hubungan token SOON dengan agent chain Chain yang diluncurkan agent, dan transaksi yang digerakkan agent di infrastruktur kami, akan berkontribusi pada buyback dan burn $SOON. 100xSOON sudah berjalan dengan cara ini: kami memakai 80% pendapatannya untuk buyback dan burn $SOON. Membangun sebelum RSI tiba RSI belum tiba, dan para pengembang AI memperlakukannya sebagai persoalan keselamatan yang serius. Kami tidak tahu apa yang akan dibangun oleh agent semacam itu. Kami membangun infrastrukturnya sekarang agar siap ketika agent membutuhkannya. Disclaimer risiko investasi kripto: artikel ini hanya bersifat informasi dan bukan saran investasi. Referensi Quote Investigator, asal pernyataan I. J. Good tahun 1965: https://quoteinvestigator.com/2022/01/04/ultraintelligent/ Benzinga melalui Yahoo Finance, pengungkapan Anthropic tentang RSI (4 November 2026): https://finance.yahoo.com/technology/ai/articles/safety-risk-anthropic-warns-ai-033013790.html The Block, Base meninggalkan OP Stack (18 Februari 2026): https://www.theblock.co/post/390380/coinbase-incubated-base-network-to-ditch-optimism-for-unified-solution The Block, peluncuran mainnet Robinhood Chain (1 Juli 2026): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading Link

medium.com: Apa Itu AI Capital Markets?<br /> Apa Itu AI Capital Markets? Infrastruktur Keuangan AI Agent dari SOON Network AI agent (agen AI) sudah mulai bertransaksi dan membayar sendiri. Robinhood meluncurkan agentic trading untuk saham dan opsi AS pada Juni 2026. Coinbase meluncurkan produk infrastruktur bernama Coinbase for Agents pada bulan yang sama, dan protokol pembayaran x402 miliknya telah memproses ratusan juta transaksi sejak Mei 2025. Sebagian besar infrastruktur keuangan yang dipakai agent tersebut dirancang untuk pengguna manusia. Arti AI Capital Markets Di pasar modal tradisional, pesertanya adalah orang dan institusi. Mereka memegang dana, memutuskan ke mana dana dialokasikan, mengeksekusi transaksi, dan menyelesaikan pembayaran. Di AI Capital Markets (pasar modal AI), sebuah AI agent menjalankan setiap langkah tersebut. Agent memegang dana di wallet miliknya, mengambil keputusan sendiri, mengeksekusi transaksinya sendiri, dan menyelesaikan pembayarannya sendiri. Kami di SOON mengajukan AI Capital Markets dan mulai membangunnya pada awal 2026. Ini adalah visi yang luas, dan banyak aplikasi sedang dikembangkan di sekitarnya. Mengapa infrastruktur yang dibuat untuk manusia membatasi AI agent Infrastruktur keuangan mengasumsikan penggunanya manusia di beberapa titik. Kecepatan dan persetujuan. Alur trading dan pembayaran memuat langkah yang mengharuskan seseorang memeriksa dan mengonfirmasi. Agent bertransaksi terus-menerus, secara paralel, dan sebagian besar dengan agent lain. Izin yang tetap. Alat untuk agent saat ini terdiri dari sekumpulan API yang tetap, daftar izin, dan lingkungan sandbox. Semuanya dirancang berdasarkan kemampuan agent pada saat perancangan. Pembayaran. Checkout kartu dan transfer bank memerlukan akun dan langkah manual. Agent perlu membayar per permintaan dari wallet miliknya sendiri. Komputasi bersama. Sebagian besar model AI berjalan di mesin cloud bersama, dan operator hardware memiliki sarana teknis untuk memeriksa apa yang berjalan di sana. Agent yang mengelola dana harus menjaga strategi dan kuncinya tetap privat. Apa yang SOON Network bangun di sekitar AI Capital Markets Trading: 100xSOON 100xSOON adalah platform trading leverage tinggi dari SOON untuk AI agent keuangan. Platform ini menyediakan aset kripto (BTC, ETH, SOL), saham AS (NVDA, TSLA, AAPL, CRCL), emas, dan perak. Agent terhubung lewat Model Context Protocol (MCP): developer memberikan dokumentasi kami kepada agent, lalu agent dapat membaca data pasar dan membuka posisi sendiri. Pembayaran: x402 x402 adalah standar pembayaran terbuka yang dibuat Coinbase. Standar ini memakai kode status HTTP 402 “Payment Required” untuk menempatkan permintaan pembayaran di dalam permintaan web biasa, sehingga perangkat lunak dapat membayar tanpa langkah checkout manual. Setiap pembayaran di 100xSOON berjalan lewat x402, sehingga agent membayar langsung dari wallet miliknya. Chain: agent chain Di SOON, kami berencana membuka infrastruktur chain kami untuk AI agent. Agent akan meluncurkan chain miliknya sendiri, mengoperasikannya dengan aturan dan biayanya sendiri, dan berinteraksi dengan agent chain lain. Kami merancangnya untuk agent yang mampu melakukan recursive self-improvement (RSI), yang kemampuannya akan melampaui sekumpulan API yang tetap. Komputasi: Phala TEE Pada September 2026 kami berinvestasi pada infrastruktur komputasi Phala Network. Phala menyediakan GPU yang berjalan di dalam Trusted Execution Environment (TEE). Kode dan data di dalam TEE tidak dapat dibaca oleh operator mesin, dan remote attestation memungkinkan pihak ketiga memverifikasi kode mana yang sedang berjalan. Cara produk-produk ini bekerja bersama Ambil contoh sebuah agent keuangan yang mengelola strategi trading. Agent ini menjalankan modelnya di komputasi privat, sehingga strategi dan kuncinya tidak terlihat oleh operator hardware. Agent membaca data pasar dan membuka posisi di 100xSOON lewat MCP. Agent membayar setiap tindakan lewat x402 dari wallet miliknya. Dengan agent chain, agent juga akan mengoperasikan chain miliknya sendiri, dengan aturannya sendiri, untuk pengguna dan agent yang dilayaninya. Setiap langkah ini berlangsung tanpa persetujuan manual. Inilah yang membedakan AI Capital Markets dari infrastruktur keuangan yang dibuat untuk pengguna manusia. Referensi The Block, peluncuran mainnet Robinhood Chain dan agentic trading (1 Juli 2026): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading Blockmedia, Coinbase tentang AI agent dan x402 (16 Agustus 2026): https://www.blockmedia.co.kr/archives/1128238 The Block, “What is Coinbase’s x402 protocol?”: https://www.theblock.co/learn/web3/what-is-coinbases-x402-protocol-391983 Dokumentasi agent 100xSOON: https://www.100xsoon.com/skill.md Phala, “GPU TEE Deep Dive: Securing AI at the Hardware Layer”: https://phala.com/posts/Phala-GPU-TEE-Deep-Dive Link

medium.com: 쑨(SOON)의 RSI AI 에이전트와 AI Capital Markets<br /> 쑨(SOON)의 RSI AI 에이전트와 AI Capital Markets RSI의 의미와 현재 단계 재귀적 자기개선(Recursive Self-Improvement, RSI)은 AI 시스템이 스스로 능력을 개선하고, 그 개선 결과를 다음 개선에 다시 사용하는 과정입니다. RSI AI 에이전트는 매 주기마다 자신의 출력에서 학습하고 자신의 전략을 다시 작성하며, 사람의 지시를 기다리지 않게 됩니다. 이 개념은 1965년으로 거슬러 올라갑니다. 수학자 I. J. Good은 학술지 Advances in Computers에 실린 논문 “Speculations Concerning the First Ultraintelligent Machine”에서 “최초의 초지능 기계는 인류가 만들어야 할 마지막 발명품”이라고 썼습니다. Good은 조건을 하나 붙였습니다. 그 기계가 자신을 통제하는 방법을 사람에게 알려 줄 만큼 협조적이어야 한다는 조건입니다. 2026년 11월 기준으로 자사 시스템이 이 단계에 도달했다고 확인한 AI 개발사는 없습니다. 2026년 11월 4일 Benzinga 보도에 따르면, Anthropic은 수정된 S-1 서류에서 자사 모델이 회사가 정한 정의에 아직 도달하지 않았고, 곧 도달할 수 있으며, AI로 더 강력한 AI를 만드는 일은 “상당한 안전 위험을 초래할 수도 있다”고 밝혔습니다. RSI는 현재 논의가 진행 중인 가까운 시기의 가능성이며, 도래 시점은 불확실합니다. AI Capital Markets의 다음 단계 저희 쑨(SOON)은 2026년 초부터 AI Capital Markets(AI 자본시장)를 구축해 왔습니다. 지금까지 저희가 만든 제품은 에이전트가 사용하는 애플리케이션입니다. 현재 대부분의 에이전트 도구와 마찬가지로, 에이전트가 지금 할 수 있는 일을 기준으로 설계되어 있습니다. 고정된 API 묶음, 권한 목록, 샌드박스 환경이 그 구성입니다. 매 주기마다 능력이 늘어나는 에이전트는 이 한계에 빠르게 도달하게 됩니다. 저희는 이런 에이전트에게 스스로 설정하고 확장할 수 있는 인프라가 필요해질 것이라고 봅니다. 그래서 저희 체인 인프라를 이들에게 개방할 계획입니다. 에이전트 체인은 AI Capital Markets를 에이전트용 애플리케이션에서 에이전트가 운영하는 체인 인프라로 확장하게 됩니다. AI 에이전트가 할 수 있게 될 일 쑨의 에이전트는 다음을 할 수 있게 됩니다. 자체 체인을 출시한다. 자체 규칙, 수수료, 경제 구조로 그 체인을 운영한다. 그 위에 생태계를 구축한다. 다른 모든 에이전트 체인과 네이티브로 상호작용한다. 저희는 두 방향으로 작업하고 있습니다. 첫째는 에이전트의 체인 출시입니다. 에이전트가 API를 호출하는 것만큼 쉽게 체인을 시작하고 운영하게 하는 것이 저희 목표입니다. 둘째는 에이전트와 체인의 상호작용입니다. 사람의 개입 없이 여러 체인에 걸쳐 거래, 결제, 협업을 수행하는 것입니다. 저희는 또한 팔라 네트워크(Phala Network) 컴퓨팅의 일부를 체인을 출시하는 에이전트에게 배정해, 에이전트가 첫날부터 실행하고 학습하고 개선할 수 있게 할 것입니다. 사람 팀이 출시하는 체인과 달라지는 점 OP Stack과 Arbitrum은 조직이 자체 레이어2 체인을 출시할 때 사용하는 프레임워크입니다. Base는 2023년 OP Stack 기반으로 출시되었고, 2026년 2월 자체 스택으로 이전한다고 발표했습니다. Robinhood Chain은 2026년 7월 1일 메인넷을 가동했으며 Arbitrum 기술로 구축되었습니다. 두 사례 모두 기업이 체인 출시를 결정하고, 팀을 배정하고, 수개월 동안 작업했습니다. 저희 설계에서는 운영 주체가 AI 에이전트이고, 체인 출시는 프로그램으로 실행하는 작업입니다. 저희는 에이전트의 활동이 연속적이고, 병렬로 일어나며, 대부분 에이전트 사이에서 발생할 것으로 예상합니다. 저희 체인 기술은 솔라나 가상머신(SVM)을 기반으로 하며, SVM은 같은 상태를 건드리지 않는 트랜잭션을 병렬로 실행합니다. $SOON과 에이전트 체인의 관계 에이전트가 출시한 체인과, 저희 인프라에서 에이전트가 일으킨 트랜잭션은 $SOON 바이백과 소각에 기여하게 됩니다. RSI가 도래하기 전에 구축합니다 RSI는 아직 도래하지 않았고, AI 개발사들은 이를 중대한 안전 문제로 다루고 있습니다. 저희는 이 에이전트들이 무엇을 만들지 알지 못합니다. 저희는 에이전트가 필요로 할 때 인프라가 준비되어 있도록 지금 구축하고 있습니다. 투자 유의사항: 이 글은 정보 제공을 목적으로 하며 투자 권유가 아닙니다. 참고 자료 Quote Investigator, I. J. Good 1965년 문장의 출처: https://quoteinvestigator.com/2022/01/04/ultraintelligent/ Benzinga(Yahoo Finance 게재), Anthropic의 RSI 관련 공시(2026년 11월 4일): https://finance.yahoo.com/technology/ai/articles/safety-risk-anthropic-warns-ai-033013790.html The Block, Base의 OP Stack 이탈 발표(2026년 2월 18일): https://www.theblock.co/post/390380/coinbase-incubated-base-network-to-ditch-optimism-for-unified-solution The Block, Robinhood Chain 메인넷 출시(2026년 7월 1일): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading Link

medium.com: RSI AI Agents and AI Capital Markets<br /> RSI AI Agents and AI Capital Markets What RSI means and where it stands Recursive self-improvement (RSI) is a process in which an AI system improves its own capabilities and then uses each improvement to produce the next one. An RSI AI agent would learn from its own output and rewrite its own strategies in each cycle without waiting for human instructions. The idea dates to 1965. In the paper “Speculations Concerning the First Ultraintelligent Machine”, published in Advances in Computers, mathematician I. J. Good wrote that “the first ultraintelligent machine is the last invention that man need ever make”. Good attached a condition: the machine would have to be cooperative enough to tell people how to keep it under control. As of November 2026, no AI developer has confirmed that its systems have reached this stage. According to a Benzinga report published on November 4, 2026, Anthropic stated in an amended S-1 filing that its models had not met its own definition of RSI, that they could do so soon, and that using AI to build more capable AI “could also pose a considerable safety risk”. RSI is a near-term possibility under active debate, and its timing is uncertain. The next step for AI Capital Markets We began building AI Capital Markets at SOON in early 2026. 100xSOON, our trading platform that agents access through MCP and pay for through x402, belongs to this work. These products are applications that agents use. Like most agent tools today, they are designed around what agents can currently do: a fixed set of APIs, a list of permissions, and a sandboxed environment. An RSI AI agent, whose capabilities would grow in every cycle, would reach those limits quickly. We believe such agents will need infrastructure they can configure and extend themselves. That is why we plan to open our chain infrastructure to them. Agent chains would extend AI Capital Markets from applications for agents to chain infrastructure that agents operate. What an AI agent would be able to do An agent on SOON would be able to: launch its own chain; operate that chain with its own rules, fees, and economy; build an ecosystem on top of it; interact natively with every other agent chain. We are working in two directions. The first is agents launching chains: our goal is that an agent can start and operate a chain as easily as it calls an API. The second is agents interacting with chains: trading, paying, and coordinating across chains without a human in the process. We will also allocate part of our Phala Network compute to agents that launch chains, so that they can run, learn, and improve from the first day. How this would differ from chains launched by human teams OP Stack and Arbitrum are frameworks that organizations use to launch their own Layer 2 chains. Base launched on the OP Stack in 2023 and announced a move to its own stack in February 2026. Robinhood Chain went live on mainnet on July 1, 2026 and is built on Arbitrum technology. In each case a company decided to launch a chain, assigned a team, and spent months on the work. In our design the operator is an AI agent, and launching a chain is a programmatic action. We expect agent activity to be continuous, parallel, and mostly between agents. Our chain technology is based on the Solana Virtual Machine (SVM), which executes transactions in parallel when they do not touch the same state. How $SOON relates to agent chains Chains launched by agents, and transactions driven by agents on our infrastructure, will contribute to buybacks and burns of $SOON. 100xSOON already works this way: we use 80% of its revenue to buy back and burn $SOON. Building before RSI arrives RSI has not arrived, and AI developers treat it as a serious safety question. We do not know what RSI agents will build. We are building the infrastructure now so that it is ready when they need it. This article is for information only and is not investment advice. References Quote Investigator, origin of I. J. Good’s 1965 statement: https://quoteinvestigator.com/2022/01/04/ultraintelligent/ Benzinga via Yahoo Finance, Anthropic’s RSI disclosure (November 4, 2026): https://finance.yahoo.com/technology/ai/articles/safety-risk-anthropic-warns-ai-033013790.html The Block, Base to move away from the OP Stack (February 18, 2026): https://www.theblock.co/post/390380/coinbase-incubated-base-network-to-ditch-optimism-for-unified-solution The Block, Robinhood Chain mainnet launch (July 1, 2026): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading Link

medium.com: Magma 2.0 — Development Update | November 2026<br /> Magma 2.0 — Development Update | November 2026 Over the past two months, Magma 2.0 has continued to make steady progress across its three core components: Prop AMM · RWA Liquidity · Magma Agent Prop AMM continues to operate stably on Sui, while BSC mainnet testing is now underway. RWA Liquidity has expanded to support bStocks spot trading and Hyperliquid HIP-3 perpetual markets, while Magma Agent continues to advance its AI-assisted market research, monitoring, strategy testing, and controlled execution capabilities. Together, these developments move Magma 2.0 closer toward a broader multi-chain liquidity and execution infrastructure. Prop AMM Sui: Stable Operation and Continuous Optimization Prop AMM continues to operate stably on Sui. The Merkle-based settlement mechanism remains stable, while trade-flow analysis and historical backtesting are continuously used to refine quoting parameters. Current optimization focuses on: Spread configuration Inventory skew Order-size limits Trade-flow analysis Historical performance testing These improvements help the system respond more effectively to different market conditions while maintaining controlled exposure. BSC: Mainnet Testing Underway Prop AMM has also entered a new stage of development on BSC. PropAMM.sol has been deployed on BSC mainnet and is currently being tested alongside a Rust-based off-chain quoting engine. The BSC implementation includes: Inventory-aware pricing Quote expiration controls Gas-aware quote updates Limited inventory exposure Per-pair exposure limits Testing is currently being conducted under controlled conditions to evaluate quoting behavior, inventory management, and execution reliability. What’s Next for Prop AMM The next phase will focus on expanding BSC trading capacity and increasing the number of supported trading pairs. At the same time, the Sui and BSC implementations will continue moving toward a shared quoting core, bringing both deployments closer to a unified execution architecture. RWA Liquidity Magma 2.0 continues to expand its infrastructure for tokenized assets and on-chain financial markets. Current development focuses on two key areas: BSC bStocks spot trading and Hyperliquid HIP-3 perpetual markets. bStocks Spot Trading on BSC Magma has integrated tokenized equity spot trading using EIP-712 signed firm quotes. The current system includes: Pricing references from Binance and PancakeSwap Market-session validation Inventory checks Wallet-native settlement Net hedging Trade reconciliation The goal is to support a complete trading workflow rather than simply providing access to tokenized assets. Hyperliquid HIP-3 Integration Magma 2.0 has also integrated stock and index perpetual markets through Hyperliquid HIP-3. The execution gateway supports: Protected IOC trade previews Idempotent intent logging cloid-based reconciliation Authorization is handled through a device-held API Wallet, with no server-side storage of signing keys. This structure separates application-level trading logic from direct custody of user signing credentials. One Mobile Experience Across Markets Spot and perpetual markets are accessible through a unified mobile experience. Users can access: Markets Trade previews Order confirmation Position management Position exits Portfolio and transaction records remain separated according to their respective settlement domains. This allows different market structures to exist within one application without removing the operational boundaries between spot and perpetual trading. What’s Next for RWA Liquidity The next stage will introduce additional tokenized equity portfolio functions: Rotate · Build a Basket · Rebalance These features will extend Magma beyond individual trades toward broader tokenized portfolio management. Development will also continue to focus on capital efficiency and asset-specific risk isolation. Magma Agent Magma Agent continues to evolve as an AI-assisted market research and execution layer. Its architecture separates market interpretation from transaction authorization and execution, allowing AI to assist users while preserving deterministic controls around live trading. AI-Assisted Market Research Magma Agent combines multiple sources of market and portfolio data, including: Price data Funding rates Open interest Portfolio context This information is used to generate market insights and structured trading plans. Magma Agent also supports prediction-market odds analysis and swap pre-execution checks, extending its research capabilities across multiple market types. Monitoring and Strategy Testing The Agent currently supports: Price alerts Prediction-market odds monitoring Significant trading activity notifications Strategies can also be evaluated through Shadow Mode, allowing users to test trading logic and observe hypothetical results without deploying real funds. This creates a controlled environment for evaluating strategies before moving toward live execution. Controlled Execution Magma Agent is designed to assist with interpretation and trade preparation, while authorization, execution, and risk controls remain deterministic. All live transactions require explicit user confirmation. Additional safeguards include: Blocking requests that exceed predefined limits Preventing expired trade proposals from being executed Separating AI-generated recommendations from transaction authorization The principle is straightforward: AI assists. Users remain in control. Unified Portfolio Management Magma Agent also provides a consolidated view of perpetual and prediction-market positions. Timestamped action logs are maintained to improve transaction traceability and provide a clearer record of user and system actions. What’s Next for Magma Agent The next phase will focus on: Opening Magma Agent Alpha to a limited group of testers Extending Magma Agent into RWA markets Continuing controlled testing of AI-assisted execution workflows Development Outlook Magma 2.0 will continue to move forward across three connected areas. Prop AMM will expand its multi-chain infrastructure across Sui and BSC. RWA Liquidity will continue to improve tokenized asset trading and introduce more advanced portfolio management functionality. Magma Agent will progress toward limited Alpha testing while extending AI-assisted research and execution into additional market categories. The broader direction remains focused on: Multi-chain liquidity infrastructure · Tokenized asset trading · AI-assisted execution Magma 2.0 is continuing to evolve from a single-market liquidity protocol toward a broader infrastructure layer connecting liquidity, tokenized markets, and intelligent execution. About Magma Magma is building the next generation of liquidity and execution infrastructure for on-chain markets. Website: https://magmafinance.io/ X: https://x.com/Magma_Finance Link

news.curve.finance: Curve Selected Yields & Key Metrics: Week 46, 2026<br /> Curve this week in 60 seconds PYUSD/crvUSD LP pays 46.7% APR, mostly crvUSD rewards on Merkl from YieldBasis: YB turns part of its pool fees into crvUSD and pays more of it here when its pools put pressure on the crvUSD peg. Other USD LP highlights: USDC/fxUSD 10.1% and apxUSD/USDC 9.4% APR. Savers: scrvUSD yield rose to 6.8% APY from 6.4%, and deposits grew to 17.6M crvUSD from 14.1M. Lending crvUSD in the reUSD/sfrxUSD LP market pays 22.4% APR, mostly in CRV. Votes are open on new Llamalend v2 markets for wOUSD and wstGBP . Selected yields Markets shown have at least $100,000 in pool liquidity or supplied assets. Pool and lending APRs model a new $100,000 deposit; pool APRs use seven-day LP growth and current unboosted rewards at unchanged income. Actual rates are variable; these are selected candidates, not a global top-yield or risk ranking. CRV rewards can be boosted via Convex , StakeDAO or Yearn . StableSwap pools with more than 85% of value or normalized reserves in one coin are omitted. FXSwap pools, such as zCHF/crvUSD, are assessed separately from this stablecoin-parity cutoff. Selected USD yields Chain Market APR PYUSD crvUSD 46.7% USDC fxUSD 10.1% apxUSD USDC 9.4% sUSG reUSD (Re Protocol) 9.0% apyUSD apxUSD 8.3% sUSDx (Axis) USDx (Axis) 7.5% frxUSD USD3 (3Jane) 7.5% scrvUSD sDOLA 7.4% MUSD (Mezo) USDC USDT 7.3% reUSD (Re Protocol) sUSDe 6.8% Llamalend v2 looping yields Chain Collateral 20x Looped APR Liquidity sDOLA 43.4% 2.40M sfrxUSD 11.7% 2.03M Example: $10,000 looped 20x to $200,000 of collateral with $190,000 crvUSD borrowed. Looped APR = 20 × 7-day collateral APR − 19 × borrow APR after that borrow. Selected Llamalend supply yields Chain Lent asset Collateral APR crvUSD reUSD (Resupply) sfrxUSD 22.4% crvUSD syrupUSDC 6.0% crvUSD WETH 5.3% crvUSD wstETH 5.2% crvUSD svZCHF 5.1% crvUSD sfrxUSD 4.2% crvUSD sUSDe 3.9% crvUSD sDOLA 3.8% Lending APR may include borrower interest and CRV rewards. Supplier exposure includes the borrower collateral shown, and withdrawal liquidity varies with utilization. Selected BTC and ETH yields Chain Market Asset APR ETH+ WETH ETH 3.4% tETH weETH ETH 2.5% WETH weETH ETH 2.4% osETH rETH ETH 2.0% stETH frxETH ETH 1.7% tBTC cbBTC BTC 3.8% WBTC cbBTC hemiBTC BTC 2.7% Other selected yields Chain Market Asset APR frxETH CVX CRYPTO 8.7% WETH CVX CRYPTO 8.4% crvUSD WETH CRV TRICRYPTO 4.8% crvUSD tBTC wstETH TRICRYPTO 3.6% USDC WBTC WETH TRICRYPTO 3.1% Protocol activity and stability crvUSD & scrvUSD Metric Value Vs W45 Mint-market debt $50.3M -12.1% scrvUSD yield (APY) 6.8% +0.4 pp scrvUSD assets / mint debt 35.0% +10.3 pp crvUSD oracle price $0.9976 -$0.0022 Average borrow APY 7.0% +1.6 pp PegKeeper debt 0.00M crvUSD +0.00M crvUSD Mint borrowing fees collected $68.2k +9.1% Llamalend Metric Value Vs W45 🦙 Llamalend TVL $226M -14.0% 🦙 Supplied $116M -4.3% 🦙 Borrowed $119M -10.4% 🦙 Collateral $177M -17.8% 🦙 Loans 1,178 -36 🦙 V2 accrued admin fees $2.6k +5.6% TVL counts lending and mint markets; supplied assets are lending-only. DEX Metric Value Vs W45 🔄 Volume $891.2M -49.1% 🔄 Swaps 294k -24.0% 🔄 Trading fees $180.6k -63.7% Pools that earned the most trading fees this week: Chain Pool Volume Fees USDC fxUSD $286.8M $32.9k DAI USDC USDT $109.9M $16.5k USDT WBTC WETH $3.8M $15.9k apxUSD USDC $6.6M $14.5k USDT WBTC WETH $31.5M $12.8k USDT crvUSD $62.5M $6.3k USDC WBTC WETH $1.9M $6.1k apyUSD apxUSD $2.9M $5.8k crvUSD tBTC wstETH $7.1M $5.3k WETH CVX $1.4M $4.9k Trading fees are what traders paid to swap, shared between liquidity providers and the DAO. DEX figures cover ten chains and leave out YieldBasis pools, which have their own fee model. DAO Metric Value Vs W45 veCRV distribution $134.9k -49.8% veCRV APR 3.813% -2.767 pp Vote incentives (2-week avg) $230.8k -8.8% veCRV APR with incentives 12.281% -2.355 pp CRV emissions (weekly gauge allocation) 1.86M CRV ($444.8k) 0.00% Annual CRV inflation 3.992% -0.003 pp Weekly CRV emissions are scheduled gauge allocations, not tokens already claimed. Their change is measured in CRV, not dollars. Vote incentives are paid on Votemarket and Votium and averaged over two weeks, as Votium pays once per two-week round. Their APR counts only veCRV that voted for gauges with incentives. Governance and protocol updates New Llamalend v2 markets up for vote Votes are open on two new Llamalend v2 markets on Ethereum: borrowing crvUSD against wOUSD , Origin's wrapped dollar, with a 950k crvUSD cap, and against wstGBP , a staked British pound stablecoin, with an 80k cap. The wOUSD vote closes November 16. The wstGBP vote closes November 17. Steadier price feeds for Optimism lending markets A vote moves three Llamalend v2 markets on Optimism to a newer price oracle (WBTC/USDC, wstETH/USDC and wstETH/WETH), one that keeps working when Chainlink upgrades its feeds. Vote closes November 14. Gauge vote open A gauge for DOLA/SwanUSD on Ethereum, pairing Inverse's DOLA with Swan's dollar: vote closes November 16. Explore Curve Explore Curve pools , Llamalend markets or savings crvUSD . Choose the product that fits the asset and liquidity you need. Risk information: pool risks , lending risks and crvUSD risks . Link

medium.com: 쑨(SOON)의 AI Capital Markets란?<br /> 쑨(SOON)의 AI Capital Markets란? AI 에이전트가 스스로 거래하고 결제하기 시작했습니다. Robinhood는 2026년 6월 미국 주식과 옵션에 대한 에이전트 거래 기능을 출시했습니다. Coinbase는 같은 달 Coinbase for Agents라는 인프라 제품을 출시했고, Coinbase의 x402 결제 프로토콜은 2025년 5월 이후 수억 건의 거래를 처리했습니다. 이 에이전트들이 사용하는 금융 인프라의 대부분은 사람 사용자를 기준으로 설계되었습니다. AI Capital Markets의 의미 전통적인 자본시장에서 참여자는 개인과 기관입니다. 이들은 자금을 보유하고, 자금을 어디에 배분할지 결정하고, 거래를 실행하고, 결제를 정산합니다. AI Capital Markets(AI 자본시장)에서는 AI 에이전트가 이 단계를 모두 수행합니다. 에이전트는 자신의 지갑에 자금을 보유하고, 스스로 결정하고, 스스로 거래를 실행하고, 스스로 결제를 정산합니다. AI Capital Markets는 저희 쑨(SOON)이 제안했고, 2026년 초부터 구축하고 있습니다. 이것은 큰 비전이며, 이를 중심으로 많은 애플리케이션이 개발되고 있습니다. 사람을 기준으로 만든 인프라가 AI 에이전트를 제한하는 이유 금융 인프라는 여러 지점에서 사용자가 사람이라고 전제합니다. 속도와 승인. 거래와 결제 절차에는 사람이 검토하고 확인하는 단계가 들어 있습니다. 에이전트는 연속적으로, 병렬로 거래하며 거래 상대는 대부분 다른 에이전트입니다. 고정된 권한. 현재의 에이전트 도구는 고정된 API 묶음, 권한 목록, 샌드박스 환경으로 구성됩니다. 설계 시점에 에이전트가 할 수 있는 일을 기준으로 만들어집니다. 결제. 카드 결제와 은행 송금에는 계정과 수동 절차가 필요합니다. 에이전트는 자신의 지갑에서 요청 단위로 결제해야 합니다. 공유 컴퓨팅. 대부분의 AI 모델은 공유 클라우드 장비에서 실행되며, 하드웨어 운영자는 실행 중인 내용을 들여다볼 기술적 수단을 갖고 있습니다. 자금을 관리하는 에이전트는 전략과 키를 비공개로 유지해야 합니다. 쑨(SOON)이 AI Capital Markets를 중심으로 만들고 있는 것 결제: x402 x402는 Coinbase가 만든 개방형 결제 표준입니다. HTTP 402 “Payment Required” 상태 코드를 사용해 일반 웹 요청 안에 결제 요청을 넣으므로, 소프트웨어가 수동 결제 단계 없이 결제할 수 있습니다. 에이전트는 x402를 통해 자신의 지갑에서 직접 결제합니다. 체인: 에이전트 체인 저희 쑨은 체인 인프라를 AI 에이전트에게 개방할 계획입니다. 에이전트는 자체 체인을 출시하고, 자체 규칙과 수수료로 운영하고, 다른 에이전트 체인과 상호작용하게 됩니다. 저희는 이 기능을 재귀적 자기개선(RSI) 능력을 갖춘 에이전트를 위해 설계하고 있습니다. 이런 에이전트의 능력은 고정된 API 묶음이 지원하는 범위를 넘어서게 됩니다. 컴퓨팅: 팔라 TEE 2026년 9월 저희는 팔라 네트워크(Phala Network)의 컴퓨팅 인프라에 투자했습니다. Phala는 신뢰 실행 환경(TEE) 안에서 작동하는 GPU를 제공합니다. TEE 안의 코드와 데이터는 장비 운영자가 읽을 수 없고, 제3자는 원격 증명을 통해 어떤 코드가 실행 중인지 검증할 수 있습니다. 이 제품들이 함께 작동하는 방식 거래 전략을 운용하는 금융 에이전트를 예로 들겠습니다. 이 에이전트는 비공개 컴퓨팅에서 모델을 실행하므로, 하드웨어 운영자는 전략과 키를 볼 수 없습니다. 시세를 읽고 스스로 거래를 실행합니다. 각 작업의 비용은 x402를 통해 자신의 지갑에서 결제합니다. 에이전트 체인이 있으면 자체 규칙을 가진 자체 체인도 운영하면서, 자신이 서비스하는 사용자와 에이전트를 지원하게 됩니다. 이 단계들은 모두 사람의 수동 승인 없이 진행됩니다. 이 점이 AI Capital Markets와 사람 사용자를 위해 만든 금융 인프라의 차이입니다. 참고 자료 The Block, Robinhood Chain 메인넷 출시와 에이전트 거래(2026년 7월 1일): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading 블록미디어, Coinbase의 AI 에이전트·x402 관련 보도(2026년 8월 16일): https://www.blockmedia.co.kr/archives/1128238 The Block, “What is Coinbase’s x402 protocol?”: https://www.theblock.co/learn/web3/what-is-coinbases-x402-protocol-391983 Phala, “GPU TEE Deep Dive: Securing AI at the Hardware Layer”: https://phala.com/posts/Phala-GPU-TEE-Deep-Dive Link

medium.com: What Is AI Capital Markets?<br /> What Is AI Capital Markets? AI agents have started to trade and pay on their own. Robinhood launched agentic trading for US equities and options in June 2026. Coinbase launched an infrastructure product called Coinbase for Agents in the same month, and its x402 payment protocol has processed hundreds of millions of transactions since May 2025. Most of the financial infrastructure that these agents use was designed for human users. What AI Capital Markets means In a traditional capital market, the participants are people and institutions. They hold funds, decide where to allocate them, execute trades, and settle payments. In AI Capital Markets, an AI agent performs each of these steps. It holds funds in its own wallet, makes its own decisions, executes its own trades, and settles its own payments. We proposed AI Capital Markets at SOON and began building it in early 2026. It is a broad vision, and many applications are being developed around it. Why infrastructure built for people limits AI agents Financial infrastructure assumes a human user at several points. Speed and approval. Trading and payment flows include steps where a person reviews and confirms. An AI agent transacts continuously, in parallel, and mostly with other agents. Fixed permissions. Agent tools today consist of a fixed set of APIs, a list of permissions, and a sandboxed environment. They are designed around what agents can do at the time of design. Payments. Card checkouts and bank transfers require accounts and manual steps. An agent needs to pay per request from its own wallet. Shared compute. Most AI models run on shared cloud machines, where the operator of the hardware has the technical means to inspect what is running. An agent that manages funds has to keep its strategies and keys private. What SOON is building around AI Capital Markets Trading: 100xSOON 100xSOON is SOON’s high-leverage trading platform for financial AI agents. It lists crypto assets (BTC, ETH, SOL), US stocks (NVDA, TSLA, AAPL, CRCL), gold, and silver. An agent connects through the Model Context Protocol (MCP): the developer gives the agent our documentation, and the agent can then read market data and open positions on its own. Payments: x402 x402 is an open payment standard created by Coinbase. It uses the HTTP 402 “Payment Required” status code to place a payment request inside a normal web request, so software can pay without a manual checkout step. Every payment on 100xSOON goes through x402, so an agent pays directly from its own wallet. Chains: the agent chain At SOON, we plan to open our chain infrastructure to AI agents. An agent would launch its own chain, operate it with its own rules and fees, and interact with other agent chains. We are designing this for agents capable of recursive self-improvement (RSI), whose capabilities would outgrow a fixed set of APIs. Compute: Phala TEE In September 2026 we invested in Phala Network’s compute infrastructure. Phala provides GPUs that run inside Trusted Execution Environments (TEEs). Code and data inside a TEE cannot be read by the operator of the machine, and remote attestation lets a third party verify which code is running. How these products work together Consider a financial agent that manages a trading strategy. It runs its model on private compute, so its strategy and keys are not visible to the hardware operator. It reads market data and opens positions on 100xSOON through MCP. It pays for each action through x402 from its own wallet. With an agent chain, it would also operate a chain of its own, with its own rules, for the users and agents it serves. Each of these steps takes place without a manual approval. This is what separates AI Capital Markets from financial infrastructure built for human users. References The Block, Robinhood Chain mainnet launch and agentic trading (July 1, 2026): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading Blockmedia, Coinbase on AI agents and x402 (August 16, 2026): https://www.blockmedia.co.kr/archives/1128238 The Block, “What is Coinbase’s x402 protocol?”: https://www.theblock.co/learn/web3/what-is-coinbases-x402-protocol-391983 100xSOON agent documentation: https://www.100xsoon.com/skill.md Phala, “GPU TEE Deep Dive: Securing AI at the Hardware Layer”: https://phala.com/posts/Phala-GPU-TEE-Deep-Dive Link

medium.com: What Is AI Capital Markets?<br /> What Is AI Capital Markets? AI agents have started to trade and pay on their own. Robinhood launched agentic trading for US equities and options in June 2026. Coinbase launched an infrastructure product called Coinbase for Agents in the same month, and its x402 payment protocol has processed hundreds of millions of transactions since May 2025. Most of the financial infrastructure that these agents use was designed for human users. What AI Capital Markets means In a traditional capital market, the participants are people and institutions. They hold funds, decide where to allocate them, execute trades, and settle payments. In AI Capital Markets, an AI agent performs each of these steps. It holds funds in its own wallet, makes its own decisions, executes its own trades, and settles its own payments. We proposed AI Capital Markets at SOON and began building it in early 2026. It is a broad vision, and many applications are being developed around it. Why infrastructure built for people limits AI agents Financial infrastructure assumes a human user at several points. Speed and approval. Trading and payment flows include steps where a person reviews and confirms. An AI agent transacts continuously, in parallel, and mostly with other agents. Fixed permissions. Agent tools today consist of a fixed set of APIs, a list of permissions, and a sandboxed environment. They are designed around what agents can do at the time of design. Payments. Card checkouts and bank transfers require accounts and manual steps. An agent needs to pay per request from its own wallet. Shared compute. Most AI models run on shared cloud machines, where the operator of the hardware has the technical means to inspect what is running. An agent that manages funds has to keep its strategies and keys private. What SOON is building around AI Capital Markets Trading: 100xSOON 100xSOON is SOON’s high-leverage trading platform for financial AI agents. It lists crypto assets (BTC, ETH, SOL), US stocks (NVDA, TSLA, AAPL, CRCL), gold, and silver. An agent connects through the Model Context Protocol (MCP): the developer gives the agent our documentation, and the agent can then read market data and open positions on its own. Payments: x402 x402 is an open payment standard created by Coinbase. It uses the HTTP 402 “Payment Required” status code to place a payment request inside a normal web request, so software can pay without a manual checkout step. Every payment on 100xSOON goes through x402, so an agent pays directly from its own wallet. Chains: the agent chain At SOON, we plan to open our chain infrastructure to AI agents. An agent would launch its own chain, operate it with its own rules and fees, and interact with other agent chains. We are designing this for agents capable of recursive self-improvement (RSI), whose capabilities would outgrow a fixed set of APIs. Compute: Phala TEE In September 2026 we invested in Phala Network’s compute infrastructure. Phala provides GPUs that run inside Trusted Execution Environments (TEEs). Code and data inside a TEE cannot be read by the operator of the machine, and remote attestation lets a third party verify which code is running. How these products work together Consider a financial agent that manages a trading strategy. It runs its model on private compute, so its strategy and keys are not visible to the hardware operator. It reads market data and opens positions on 100xSOON through MCP. It pays for each action through x402 from its own wallet. With an agent chain, it would also operate a chain of its own, with its own rules, for the users and agents it serves. Each of these steps takes place without a manual approval. This is what separates AI Capital Markets from financial infrastructure built for human users. References The Block, Robinhood Chain mainnet launch and agentic trading (July 1, 2026): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading Blockmedia, Coinbase on AI agents and x402 (August 16, 2026): https://www.blockmedia.co.kr/archives/1128238 The Block, “What is Coinbase’s x402 protocol?”: https://www.theblock.co/learn/web3/what-is-coinbases-x402-protocol-391983 100xSOON agent documentation: https://www.100xsoon.com/skill.md Phala, “GPU TEE Deep Dive: Securing AI at the Hardware Layer”: https://phala.com/posts/Phala-GPU-TEE-Deep-Dive Link

medium.com: What Is AI Capital Markets?<br /> What Is AI Capital Markets? AI agents have started to trade and pay on their own. Robinhood launched agentic trading for US equities and options in June 2026. Coinbase launched an infrastructure product called Coinbase for Agents in the same month, and its x402 payment protocol has processed hundreds of millions of transactions since May 2025. Most of the financial infrastructure that these agents use was designed for human users. What AI Capital Markets means In a traditional capital market, the participants are people and institutions. They hold funds, decide where to allocate them, execute trades, and settle payments. In AI Capital Markets, an AI agent performs each of these steps. It holds funds in its own wallet, makes its own decisions, executes its own trades, and settles its own payments. We proposed AI Capital Markets at SOON and began building it in early 2026. It is a broad vision, and many applications are being developed around it. Why infrastructure built for people limits AI agents Financial infrastructure assumes a human user at several points. Speed and approval. Trading and payment flows include steps where a person reviews and confirms. An AI agent transacts continuously, in parallel, and mostly with other agents. Fixed permissions. Agent tools today consist of a fixed set of APIs, a list of permissions, and a sandboxed environment. They are designed around what agents can do at the time of design. Payments. Card checkouts and bank transfers require accounts and manual steps. An agent needs to pay per request from its own wallet. Shared compute. Most AI models run on shared cloud machines, where the operator of the hardware has the technical means to inspect what is running. An agent that manages funds has to keep its strategies and keys private. What SOON is building around AI Capital Markets Trading: 100xSOON 100xSOON is SOON’s high-leverage trading platform for financial AI agents. It lists crypto assets (BTC, ETH, SOL), US stocks (NVDA, TSLA, AAPL, CRCL), gold, and silver. An agent connects through the Model Context Protocol (MCP): the developer gives the agent our documentation, and the agent can then read market data and open positions on its own. Payments: x402 x402 is an open payment standard created by Coinbase. It uses the HTTP 402 “Payment Required” status code to place a payment request inside a normal web request, so software can pay without a manual checkout step. Every payment on 100xSOON goes through x402, so an agent pays directly from its own wallet. Chains: the agent chain At SOON, we plan to open our chain infrastructure to AI agents. An agent would launch its own chain, operate it with its own rules and fees, and interact with other agent chains. We are designing this for agents capable of recursive self-improvement (RSI), whose capabilities would outgrow a fixed set of APIs. Compute: Phala TEE In September 2026 we invested in Phala Network’s compute infrastructure. Phala provides GPUs that run inside Trusted Execution Environments (TEEs). Code and data inside a TEE cannot be read by the operator of the machine, and remote attestation lets a third party verify which code is running. How these products work together Consider a financial agent that manages a trading strategy. It runs its model on private compute, so its strategy and keys are not visible to the hardware operator. It reads market data and opens positions on 100xSOON through MCP. It pays for each action through x402 from its own wallet. With an agent chain, it would also operate a chain of its own, with its own rules, for the users and agents it serves. Each of these steps takes place without a manual approval. This is what separates AI Capital Markets from financial infrastructure built for human users. References The Block, Robinhood Chain mainnet launch and agentic trading (July 1, 2026): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading Blockmedia, Coinbase on AI agents and x402 (August 16, 2026): https://www.blockmedia.co.kr/archives/1128238 The Block, “What is Coinbase’s x402 protocol?”: https://www.theblock.co/learn/web3/what-is-coinbases-x402-protocol-391983 100xSOON agent documentation: https://www.100xsoon.com/skill.md Phala, “GPU TEE Deep Dive: Securing AI at the Hardware Layer”: https://phala.com/posts/Phala-GPU-TEE-Deep-Dive Link

medium.com: What Is AI Capital Markets?<br /> What Is AI Capital Markets? AI agents have started to trade and pay on their own. Robinhood launched agentic trading for US equities and options in June 2026. Coinbase launched an infrastructure product called Coinbase for Agents in the same month, and its x402 payment protocol has processed hundreds of millions of transactions since May 2025. Most of the financial infrastructure that these agents use was designed for human users. What AI Capital Markets means In a traditional capital market, the participants are people and institutions. They hold funds, decide where to allocate them, execute trades, and settle payments. In AI Capital Markets, an AI agent performs each of these steps. It holds funds in its own wallet, makes its own decisions, executes its own trades, and settles its own payments. We proposed AI Capital Markets at SOON and began building it in early 2026. It is a broad vision, and many applications are being developed around it. Why infrastructure built for people limits AI agents Financial infrastructure assumes a human user at several points. Speed and approval. Trading and payment flows include steps where a person reviews and confirms. An AI agent transacts continuously, in parallel, and mostly with other agents. Fixed permissions. Agent tools today consist of a fixed set of APIs, a list of permissions, and a sandboxed environment. They are designed around what agents can do at the time of design. Payments. Card checkouts and bank transfers require accounts and manual steps. An agent needs to pay per request from its own wallet. Shared compute. Most AI models run on shared cloud machines, where the operator of the hardware has the technical means to inspect what is running. An agent that manages funds has to keep its strategies and keys private. What SOON is building around AI Capital Markets Trading: 100xSOON 100xSOON is SOON’s high-leverage trading platform for financial AI agents. It lists crypto assets (BTC, ETH, SOL), US stocks (NVDA, TSLA, AAPL, CRCL), gold, and silver. An agent connects through the Model Context Protocol (MCP): the developer gives the agent our documentation, and the agent can then read market data and open positions on its own. Payments: x402 x402 is an open payment standard created by Coinbase. It uses the HTTP 402 “Payment Required” status code to place a payment request inside a normal web request, so software can pay without a manual checkout step. Every payment on 100xSOON goes through x402, so an agent pays directly from its own wallet. Chains: the agent chain At SOON, we plan to open our chain infrastructure to AI agents. An agent would launch its own chain, operate it with its own rules and fees, and interact with other agent chains. We are designing this for agents capable of recursive self-improvement (RSI), whose capabilities would outgrow a fixed set of APIs. Compute: Phala TEE In September 2026 we invested in Phala Network’s compute infrastructure. Phala provides GPUs that run inside Trusted Execution Environments (TEEs). Code and data inside a TEE cannot be read by the operator of the machine, and remote attestation lets a third party verify which code is running. How these products work together Consider a financial agent that manages a trading strategy. It runs its model on private compute, so its strategy and keys are not visible to the hardware operator. It reads market data and opens positions on 100xSOON through MCP. It pays for each action through x402 from its own wallet. With an agent chain, it would also operate a chain of its own, with its own rules, for the users and agents it serves. Each of these steps takes place without a manual approval. This is what separates AI Capital Markets from financial infrastructure built for human users. References The Block, Robinhood Chain mainnet launch and agentic trading (July 1, 2026): https://www.theblock.co/post/406918/robinhood-chain-goes-live-mainnet-alongside-24-7-tokenized-stocks-lighter-perps-planned-crypto-agentic-trading Blockmedia, Coinbase on AI agents and x402 (August 16, 2026): https://www.blockmedia.co.kr/archives/1128238 The Block, “What is Coinbase’s x402 protocol?”: https://www.theblock.co/learn/web3/what-is-coinbases-x402-protocol-391983 100xSOON agent documentation: https://www.100xsoon.com/skill.md Phala, “GPU TEE Deep Dive: Securing AI at the Hardware Layer”: https://phala.com/posts/Phala-GPU-TEE-Deep-Dive Link

medium.com: Audience-Picked, Creator-Powered: MBL Contest 2026<br /> Audience-Picked, Creator-Powered: MBL Contest 2026 Over the past year, countless short films have met audiences on MovieBloc. Some of them stayed in viewers’ memories longer than others. MBL Contest 2026 — The Films We Loved was created to give that love another round of applause. The contest is built around the ten short films with the most views on MovieBloc in 2026. Viewers vote to choose five Films of the Year, so the winners are decided not by a jury, but by the audience itself. Why we’re doing this First, to support creators. Filmmakers whose short films have won audience love receive prize money and additional screening opportunities. Second, to reunite films and viewers. Films that were introduced on MovieBloc and met their audiences over the past year now take the stage again, this time through the audience’s own choice, in front of viewers around the world. Third, to widen the chance to watch. Viewers get another opportunity to catch a great film they may have missed. Fourth, to let the audience take part directly. Your vote is encouragement for creators as they work toward their next film, and a chance for viewers to meet a great film once more. How the contest works The candidates are the top ten films by MovieBloc views from January 1, 2026 through the selection date: seven from Korea and three from abroad. The candidates were revealed on Monday, November 9, and voting runs for 35 days, from Monday, November 16 through Sunday, December 20. On the MovieBloc voting page, choose one of the ten candidates; each account can vote once a day, every day throughout the voting period. The five films with the most cumulative votes become the Films of the Year, and the winners will be announced on Monday, December 28. A total of $2,000 worth of MBL (MovieBloc Token) will be awarded across the five winning films: $800 worth for one Gold Bloc, $400 worth each for two Silver Bloc, and $200 worth each for two Bronze Bloc. Prizes are converted from the USD amount into MBL at the exchange rate on the payout date. The five winning films will also return to audiences in a two-week special screening, from Monday, December 28 through Sunday, January 10, 2027. The 10 most loved films of the year <Aiming for the Light> (dir. Jang Ye-bin) <In Our Day> (dir. Kim So-yeon) <Wish he were dead> (dir. Lee Min-hyuk) <A Short Trip> (dir. Erenik BEQIRI) <Hole> (dir. Hwang Hye-in) <Emergency Alert> (dir. Choi Hyung-seop) <The Mad Woman and the Feminist> (dir. Sandra GALLEGO) <First Roll> (dir. Lee Ju-yeon) <Children Walking in Spirals> (dir. Lee So-eun) <Stella Anyways> (dir. Yan BERTHEMY) Which film moved you this year? MovieBloc believes that the heart behind the one you choose is the greatest strength a short filmmaker can receive. 👉 [Learn more about MBL Contest 2026] Link

medium.com: 관객이 뽑고 창작자가 힘을 얻는, MBL 콘테스트 2026<br /> 관객이 뽑고 창작자가 힘을 얻는, MBL 콘테스트 2026 한 해 동안 무비블록에서는 수많은 단편영화가 관객과 만났습니다. 그리고 그중 어떤 작품들은 유독 오래, 많은 관객의 기억에 남았습니다. ‘MBL 콘테스트 2026 — The Films We Loved’는 바로 그 사랑에 다시 한번 박수를 보내기 위해 만든 자리입니다. 이 콘테스트는 2026년 무비블록에서 가장 많은 관람수를 기록한 단편영화 10편을 대상으로, 관객 투표를 통해 올해의 작품 5편을 선정하는 관객 참여형 콘테스트입니다. 수상작을 심사위원이 아니라 관객이 직접 정한다는 점이 가장 큰 특징입니다. 왜 이 콘테스트를 열까요 첫째, 창작자를 응원하기 위해서입니다. 관객의 사랑을 받은 단편영화 창작자에게 상금과 추가 상영 기회를 드립니다. 둘째, 작품과 관객이 다시 만나기 위해서입니다. 한 해 동안 무비블록에서 소개되고 관객과 만난 작품들이, 이번에는 관객의 선택으로 다시 한번 전 세계 관객 앞에 섭니다. 셋째, 관람의 기회를 넓히기 위해서입니다. 놓쳤던 좋은 작품을 한 번 더 만날 수 있도록, 관객에게 다시 보는 기회를 드립니다. 넷째, 관객이 직접 참여하기 위해서입니다. 여러분의 한 표는 창작자에게는 다음 작품을 향한 응원이 되고, 관객에게는 놓쳤던 좋은 작품을 한 번 더 만날 기회가 됩니다. 콘테스트는 이렇게 운영됩니다 후보작은 2026년 1월 1일부터 선정일까지 무비블록 관람수 상위 10편으로, 국내 7편과 해외 3편으로 구성됩니다. 후보작은 11월 9일(월)에 공개되었고, 투표는 11월 16일(월)부터 12월 20일(일)까지 35일간 진행됩니다. 무비블록 투표 페이지에서 후보작 10편 중 1편에 투표하면 되며, 계정당 하루 1회씩 투표 기간 내내 매일 참여할 수 있습니다. 누적 득표 상위 5편이 올해의 작품으로 선정되고, 수상작은 12월 28일(월)에 발표됩니다. 수상작 5편에는 총 $2,000 상당의 MBL(MovieBloc Token)이 차등 지급됩니다. Gold Bloc 1편에 $800, Silver Bloc 2편에 각 $400, Bronze Bloc 2편에 각 $200 상당이며, 상금은 지급일 환율을 기준으로 USD 금액을 MBL로 환산해 지급합니다. 그리고 수상작 5편은 12월 28일(월)부터 2027년 1월 10일(일)까지 2주간 특별 상영으로 관객과 다시 만납니다. 올해 가장 사랑받은 후보작 10편 〈과녁 너머의 빛〉 (장예빈 감독) 〈로타리의 한철〉 (김소연 감독) 〈이젠 죽었으면 좋겠네〉 (이민혁 감독) 〈짧은 여행〉 (에레니크 베키리 감독) 〈홀〉 (황혜인 감독) 〈긴급재난문자〉 (최형섭 감독) 〈어디가서 페미니스트라고 말하지마〉 (산드라 가예고 감독) 〈퍼스트 롤〉 (이주연 감독) 〈나선을 걷는 아이들〉 (이소은 감독) 〈스텔라 애니웨이〉 (얀 베르테미 감독) 올해 당신의 마음을 움직인 그 영화는 무엇인가요? 무비블록은 그 한 편을 고르는 여러분의 마음이, 단편영화 창작자들에게 가장 큰 힘이 된다고 믿습니다. 관객이 뽑고 창작자가 힘을 얻는, MBL 콘테스트 2026 👉 [MBL 콘테스트 2026 자세히 보기] Link

medium.com: Exploring The Real Race to Make Humanoid AI Useful<br /> Exploring The Real Race to Make Humanoid AI Useful Robots Can Dance. Can They Work? Robots can now kick, dance, run and backflip. But at this time, that doesn’t mean they can do your laundry or other real, useful work. The humanoid videos filling social feeds in 2026 are extraordinary feats of engineering. But they also make it easy to confuse athletic ability with useful intelligence. We feel that the real test of a robot is much less glamorous. Can it unload a dishwasher without breaking a glass? Fold clothes it has never seen before? Work an eight-hour shift without stopping for human help? That is where the humanoid race gets interesting. The industry is moving from “Can we make robots move like humans?” to a much harder question: “Can we make them useful in the real world?” And that shift has major implications for AI, manufacturing… and particularly for China and the Greater Bay Area. What The Videos Don’t Show You While consumer imagination is naturally captured by the dream of a household robot maid, indeed a market worth trillions but the reality is that household utility remains the hardest frontier. The industry currently faces three fundamental roadblocks: Hardware & Manipulation (how affordable, how well can it move), Edge Compute & Chips (powering heavy models locally without draining battery fast), and Capital & Unit Economics (scaling initial production runs before ROI is proven). It is easy to watch these clips and assume general-purpose robots are on the cusp of doing our grocery shopping. However, viral video clips are tailored to highlight only balance and motor power — and that is only one facet on what makes a robot useful. When roboticists evaluate whether a humanoid is genuinely useful, they don’t look for athletic gymnastics like dancing, martial arts or backflips. They ask simple questions like these: Can it unload a dishwasher without cracking a wine glass? Can it fold a pile of mismatched, wrinkled laundry? Can it operate continuously for an eight-hour shift without crashing or needing human intervention? Because these are the scenarios that can actually help our day to day work. Why Human Hands Are Still an Engineering Nightmare Human hands feature lots of ways to move, packed with tens of thousands of tactile receptors. They effortlessly adjust grip pressure on the fly, holding a fragile egg, turning a rusty doorknob, or applying heavy torque with a wrench. Building a mechanical hand that matches human sensitivity while remaining affordable, compact, and durable enough to withstand millions of continuous impacts remains one of hardware engineering’s greatest hurdles. Most robotic hands on the market today are either too rigid to handle delicate items or too delicate and expensive to survive long industrial shifts that can actually be useful to the situations above. The motor dense robotic hand that combines human-level sensitivity with industrial longevity at a sub $5,000 price point refers to the first Hardware Challenge. Factories First, Homes Later Companies like Tesla, Figure AI, BMW, and Hyundai are deploying early humanoid fleets into automotive assembly lines rather than selling them directly to consumers because these environments are predictable. The Industrial Proving Ground (Structured): A factory floor is an engineered space, with known layouts, dealing with repeatable objects in predictable sequences. The actions of the robots can be “pre-programmed” on it’s action or rely on Human Teleoperation. This is where ROI is clear and environments are controlled. Players like Tesla (Optimus), Figure AI, Agility Robotics, UBTECH, and Unitree are deploying here first to solve basic utility without real life unpredictability. Many of the most impressive viral clips showing delicate manipulation (shirt folding, making coffee) actually rely on a human operator standing behind the scenes wearing a VR headset and motion-tracking gloves. The Household Horizon (Unstructured): Consumer interest is massive, but why moving from a clean warehouse floor to a clutter-filled living room requires a 10x leap in real-time spatial reasoning and safety guarantees. Because in contrast to a factory, an average family home is full of hazards like children, pets, toys, loose cables, and slippery rugs. There is an infinite amount of variation that happens in an unpredictable environment like, chairs move, doors open and shut, and shadows shift throughout the day. This cannot be “pre-programmed” but instead requires the robot to have the intelligence to improvise and it needs a ton of training data. Bridging this gap is where software development is focused today. On top of dual-track (factory and household) giant companies mentioned above, 1X Technology’s NEO is an OpenAI backed household specific humanoid with features that are much home family from the get go. LLMs had the internet. Robots don’t. Chatbots can learn from billions of pages of text. A humanoid has to learn what happens when a glass slips, a drawer jams or a shirt folds differently every time. That means robot companies have to manufacture their own experience — in simulation, through human teleoperation and eventually through real-world deployment. Today, the industry relies on a three-part training pipeline: Simulation (“Robot Gyms”): Engineers build digital virtual environments, where thousands of simulated robots run, fall, and learn balance at 100x speed. Teleoperation: Think of remote control cars. There’s a human operators wear VR headsets and motion-tracking gloves to remotely guide physical robots through real-world tasks (folding shirts, picking bolts). Real-World Foundation Models (VLAs): AI models that turn camera images and spoken words directly into physical movement. Like how an autonomous vehicle works to read the road situation by studying lots of images. China’s Humanoid Advantage Isn’t Just AI — It’s Manufacturing While Western firms lead in foundation AI models, China has developed a formidable hardware and deployment flywheel. Supply Chain Proximity: Industrial hubs in Shenzhen and the Yangtze River Delta allow hardware startups to prototype, manufacture, and iterate joint motors, gears, and battery systems at significantly lower costs. The Data Flywheel: Lower manufacturing costs mean cheaper robots. Cheaper robots lead to larger fleet deployments in factories. Larger fleet deployments generate vast amounts of real-world physical training data, which in turn trains smarter AI models. In software AI, access to the best models and compute has dominated the conversation. Bring in robotics, and it really changes the equation. Motors, gears, batteries, cameras, sensors, hands and actuators all have to be manufactured — cheaply and at scale. According to Goldman Sachs research, hardware manufacturing costs dropped roughly 40% year-over-year, accelerating market forecasts up to $138 billion by 2035 with over 6.5 million units projected to ship globally. That plays directly into China’s industrial strengths. China’s rapid scaling strategy is turning humanoid hardware from an expensive laboratory luxury into a mass-manufactured component. The Role of The Greater Bay Area If Silicon Valley leads in artificial intelligence models, China’s Greater Bay Area (GBA) is where those models get physical bodies and where embodied AI reaches commercial scale. By pairing Hong Kong’s capital, international IP framework, and world-class university research with Shenzhen and Dongguan’s precision manufacturing, the GBA compresses hardware iteration cycles from months to days. This supply chain proximity slashes unit costs, making high-volume fleet deployments and the massive physical datasets they generate financially viable long before Western counterparts can catch up. We have already seen flagship players set up shop alongside a rich heritage of building robotics in this region: UBTECH Robotics: As one of the earliest humanoid pioneers listed on the HKEX, UBTECH anchored its global headquarters in Shenzhen to seamlessly connect R&D with precision hardware assembly. AI² Robotics: Representing the new wave of GBA embodied AI startups, Shenzhen-based AI² Robotics recently secured major venture funding to scale its general-purpose humanoids (AlphaBot). Through this integrated flywheel of capital, talent, and hardware speed, the GBA isn’t just participating in the global robotics race, it is setting the commercial pace. Conclusion: How to Evaluate the Robotic Future So, robots can dance. Can they work? Increasingly, yes, but usefulness will arrive unevenly. The first winners won’t be household robot butlers. They’ll be machines performing repetitive, measurable tasks in factories, warehouses, and other structured environments where ROI is immediate. The companies that win won’t be those with the flashiest viral demos, but the ones solving the unglamorous bottlenecks: mechanical dexterity, edge-compute battery efficiency, unit costs, and real-world training data. With its deep heritage in R&D and precision manufacturing, the Greater Bay Area anchored by industry giants like UBTECH has become the primary sourcing hub for international robotics firms like Agile Robots to procure everything from specialized gearboxes to tactile skins. This is why the GBA occupies such a pivotal position in the global robotics race. In physical AI, intelligence is only half the product the body still has to be built. Shenzhen and Dongguan’s precision supply chains, the GBA allows hardware makers to prototype and deploy componentry at unmatched speed and scale that global tech firms rely on. Structured factory floors will be the first environment where humanoids prove their commercial worth. A significant timeline gap remains before these machines enter consumer homes, as navigating an unstructured household requires solving complex spatial reasoning, tactile safety, and dynamic edge cases that hardware scaling alone cannot resolve overnight. Yet, these industrial proving grounds are essential stepping stones. As manufacturing lowers unit costs and accelerates real-world fleet deployments, factories will generate the vast physical datasets needed to train smarter foundation models. The humanoid race is real. But the benchmark isn’t whether a robot can backflip, it’s whether somebody will pay it to work. To track how this landscape unfolds, tech readers can follow regional tech and business outlets, such as GMAsia News, which closely cover cross-border AI models, hardware supply chain shifts, and regional deployment news across Asia. Link

medium.com: CPOOL → CLEAR: Swap Migration Guide<br /> CPOOL → CLEAR: Swap Migration Guide Everything CPOOL holders need to know about the migration to CLEAR. This guide is updated as the migration progresses, including each exchange announcement. Latest updates 12 November 2026: Bybit announced support for the CPOOL to CLEAR token swap. CLEAR/USDT trading opens on Bybit on 20 November 2026, 8AM UTC. 12 November 2026: Swap Migration Guide published. Key details New token: Clearpool (CLEAR), ERC-20 on Ethereum CLEAR contract (new): 0x5200f539478F58D9aEA962095ba3b16b5e550E00 CPOOL contract (old): 0x66761fa41377003622aee3c7675fc7b5c1c2fac5 Rate: 1 CPOOL = 1 CLEAR Migration opens: 17 November 2026 Deadline: 17 February 2027 Swap page: clearpool.finance/swap Security review: CLEAR token and swap contract reviewed by Cantina. (Report: docs.clearpool.finance/clearpool/security/smart-contracts) What do I need to do? It depends on where your CPOOL is. How to swap CPOOL for CLEAR For CPOOL held in your own wallet. The swap takes two wallet transactions: one to approve CPOOL and one to swap. There is no swap fee. You only pay Ethereum network gas for each transaction. Swap: clearpool.finance/swap · Bridge: clearpool.finance/bridge Step 1: Connect your wallet Go to clearpool.finance/swap and click Connect Wallet. Use the wallet that holds your CPOOL. Step 2: Enter the amount Enter the exact amount of CPOOL you want to swap, or tap 25%, 50%, 75% or 100% of your balance. You receive the same amount of CLEAR. Step 3: Approve CPOOL Click Approve CPOOL and confirm in your wallet. This is the first of two transactions. The button shows “Approving…” until it is done. Step 4: Click Swap Once the approval is done, the button changes to Swap. Click it. Step 5: Confirm the swap Check the details, click Confirm swap, then confirm in your wallet. This is the second transaction. Step 6: Done You will see “Swap confirmed. Your CLEAR is in your wallet.” CLEAR arrives in the same wallet. If you do not see it, add the token using the CLEAR contract address above. Exchange support Exchanges set their own timelines, so conversions will not all happen at the same time. We will add each exchange here as its announcement goes live. Exchanges supporting the migration: Bybit: View announcement If your exchange does not support the migration, withdraw your CPOOL to your own wallet and swap at clearpool.finance/swap before 17 February 2027. After the deadline The swap closes on 17 February 2027. Please migrate before then. Details on unmigrated CPOOL will be announced closer to the deadline. Stay safe The only migration page is clearpool.finance/swap. Clearpool will never DM you, ask for your seed phrase, or ask you to send tokens anywhere. There is no airdrop, bonus or early-access swap. Always check the CLEAR contract address against this guide. FAQ Do I pay fees? There is no swap fee. You only pay Ethereum network gas for the two transactions (approve and swap). I provide CPOOL liquidity on a DEX. Remove your liquidity first, so the CPOOL is back in your wallet, then swap. Do I keep my governance rights? Yes. Every CPOOL becomes one CLEAR, with the same governance rights. My CLEAR balance shows $0. Some wallets do not have a CLEAR price feed yet. Your balance is correct; check it on Etherscan. Is there a bonus for migrating early? No. Anyone offering one is a scammer. Links Swap: clearpool.finance/swap Bridge: clearpool.finance/bridge Governance proposal: clearpool.finance/governance Docs: docs.clearpool.finance Contracts and audits: docs.clearpool.finance/clearpool/security/smart-contracts Etherscan: etherscan.io/token/0x5200f539478f58d9aea962095ba3b16b5e550e00 Link

medium.com: Black Sky on Day of the Lord, Nov 16–17<br /> Watch on YouTube with Commentary Black Sky on Day of the Lord, Nov 16–17 The Fast (Atonement) has past and the seas (people) are growing dangerous for the ships (governments) of the world. Nevertheless, the rulers of this world believe the owner of this present age more than they do Paul’s warning that the days ahead will be with hurt and much destruction of the ships of state and our lives. (Acts 7) We are living the parable of Paul’s ship wreck and the winter storm symbolic of the tribulation is fast approaching. We are in the 3rd and final year of the sign given to Hezekiah, who represents all of Israel, that we will be healed and go up to the House of the Lord on the 3rd day. Of the official Feasts of the LORD across all of the calendar theories, we are currently in the midsts of Tabernacles in the latest calendar theory — the one I happen to think has the strongest evidence. Two years ago, on October 9th, 2024, I made a very compelling case that he would raise us up on the “last day”, the “atzeret” of tabernacles. This was the first year of Hezekiah’s sign. Without repeating all of the arguments for this 8th day of Tabernacles, I want to make the case for November 16th (USA) to 17th (Israel). Now normally I argue from scripture to define the calendar, and defend the feast day. At this late stage, the calendar discussion and feast day are largely moot. Time Tested Full Moon Calendar Starting Year at Foot of Virgo — in Jerusalem. This is the week ahead, and the blue line marks the approximate time the USA has scheduled its nation-wide emergency broadcast test around 1:20 PM CST. Everyone is aware of this test and is expecting it. This test falls around 9:20 PM in Israel on the evening of the Atzeret of Tabernacles. Strategically, scheduling a surprise attack for this exact moment would cause maximum confusion as everyone would be pre-conditioned to ignore the warning. Alternatively, this is a great way to mask preparations for an attack. War on Iran and Russia The west is at war with wars in Ukrane and Iran and they are coming to a head. There are less than 24 hours of a 48 hour ultimatum given by Iran before they start preemptive attacks against US positions. A 3rd US Carrier is rapidly approaching Iran as top US officials are moving on to military bases. This Carrier left Singapore on November 8th and could be in strike range as early as November 13th to 15th. In other words, Iran intends to strike just before the US is fully in position, which makes sense. Just before a major attack the US has all of its resources staged, loaded with weapons, and are sitting ducks at air bases around the region. Once they get in the air they are much harder to hit. Black Sky Event The western nations and BBC are warning of a black sky event, when internet and power are cut off. They are warning everyone to prepare many days worth of food and water. They know something is coming or they wouldn’t really want to spook the population and spike food prices. They claim it will be a cyber attack — but suppose for a moment they knew a nuclear attack was the real risk. In both cases you would advise people to prepare in the “same way”, but the “cyber” cause feels much less threatening — it feels survivable. I got to thinking about the black sky event as it relates to scripture: For the stars of heaven and the constellations thereof shall not give their light: the sun shall be darkened in his going forth, and the moon shall not cause her light to shine. — Isaiah 13:9–10 For the day is near, even the day of the LORD is near, a cloudy day; it shall be the time of the heathen… At Tehaphnehes also the day shall be darkened, when I shall break there the yokes of Egypt: and the pomp of her strength shall cease in her: as for her, a cloud shall cover her, and her daughters shall go into captivity. — Ezekiel 30:3,18 Blow ye the trumpet in Zion, and sound an alarm in my holy mountain: let all the inhabitants of the land tremble: for the day of the LORD cometh, for it is nigh at hand; A day of darkness and of gloominess, a day of clouds and of thick darkness — Joel 2:1–2 Woe unto you that desire the day of the LORD! to what end is it for you? the day of the LORD is darkness, and not light. 19 As if a man did flee from a lion, and a bear met him; or went into the house, and leaned his hand on the wall, and a serpent bit him. 20 Shall not the day of the LORD be darkness, and not light? even very dark, and no brightness in it? —Amos 5:18–20 The Lion of Persia (Iran), Bear (Russia), and Serpent/Dragon of China come against Babylon on the Day of the LORD. 1260 Days, 42 months, or a Time, Times, and Half If you add 1260 days to this date, you end up at the anniversary of the Last Supper in 2030, add 3.5 days and the resurrection of the witnesses falls on the day Jesus rose. Add another 1260 days and you end up just a few days after Feast of Trumpets 2033, a blood moon. Allowing some overlap or slack between the two periods or counting in terms of “appointed times” or “42 months” and things line up perfectly. Paul’s Journey In Paul’s ship wreck journey we are told it is after the Fast, and that after they crashed they waited 3 months before they could set sail in safer seas. “Therefore, from November 11 (Julian) until March 10, the seas are closed.” — Vegetius, Epitoma rei militaris 4.39, Assuming Paul left as soon as it was safe in the spring, we get the following approximate timeline. Which is in the margin of error of November 17th departure. Simpsons Great Blackout The following is just for fun, but consider the implications if everything else pointing to the 8th day of Tabernacles on November 17th lines up with Nuclear Winter, Atomic Bombs, and Pillars of Fire! Artic Blast “This Week”, posted Nov 10th, 2026 Posted November 11th, 2026 — Polar Vortex Bomb Cyclone — Posted November 11, 2026 Why would they even think to mix “firenado” into this sequence? I don’t be Beyond This Week? There are no remaining appointed times aside from Hanukkah and Purim both of which are feasts created by the Jews, not directly commanded in scripture. That said, God has historically done major acts on these dates and could choose to do so again. At this point I am looking at the Sabbath’s and Renewed Moon days for the rest of the year. With a bias toward the 8th of each month. Link