Palantir co-founder focuses on crypto-stock meme coin MONITOR, which has surged six-fold.

Beating AI News Flash: Chinese AI startup StepFun has open-sourced its AI coding tool Step Code v0.1.0 under the MIT license. The tool runs directly in the terminal, supporting code reading and writing, test execution, and development task handling, with compatibility for MCP, Agent Skills, and Claude Code plugins. Its built-in StepPage feature allows users to publish local static websites online with a single command. Step Code’s core highlight is low token consumption. In StepFun’s self-conducted Terminal-Bench 2.1 benchmark, it completed 72 out of 89 tasks, scoring 80.9% – tying DeepSeek Harness for the highest score while using fewer tokens. In StepFun’s in-house Multi-Frame long-task test (150 questions total), Step Code finished 110 tasks for a 73.3% success rate, averaging 5.09 million tokens per task. Among the 6 tested harnesses, it ranked first for both highest task completion rate and lowest token usage. However, StepFun did not disclose the specific underlying model used for the benchmark, nor did it detail how token consumption was reduced. The company only noted that Step Code was optimized for token efficiency alongside its Step series models. Numerous token-saving designs are visible in the source code: file reads and command outputs are length-limited, search results are narrowed as much as possible, long conversations default to context compression, and the system prompt prioritizes direct tool calls over spawning additional agents. There is also a more aggressive contextProjection feature that trims old tool outputs, duplicate results, and historical reasoning, though it is disabled by default. Overall, Step Code’s token advantage comes from multiple optimizations to both the model and the test harness.

Ethereum co-founder Vitalik Buterin delivered a keynote speech at the 12th Global Blockchain Summit, stating that while early blockchain boasted features like openness, neutrality, and security, it faced critical flaws including poor scalability, full transparency, and inadequate privacy protection. Zero-knowledge proofs— which enable private transactions and scalable verification— have moved from planning to actual Ethereum Improvement Proposals (EIPs), with technologies such as fully homomorphic encryption also seeing initial applications. Indistinguishability obfuscation may find on-chain use in the future, he added. Parallel block construction and verification can deliver higher scalability, and the Web3 industry’s focus has expanded from digital assets to both digital assets and information, covering the full workflow spanning user ends, transaction mempools, and on-chain processes. With AI advancement, feeding cryptographic rules to AI for automated formal verification can boost verification efficiency.

According to an official announcement, Binance will gradually migrate crypto assets from users’ Funding Accounts to Spot Accounts starting September 29, 2026, in a move to simplify its user account structure and enhance asset management experience. The migration is expected to run through January 2027, with specific dates to be announced separately. After the migration period concludes, the Funding Account will be renamed the Stocks Account, which will thereafter be used exclusively for stock and stock option trading and settlement. Key details include: Starting September 29, 2026, the Funding Account will no longer support on-chain deposits. Users can trade stocks and stock options via the Stocks Account, which supports six settlement assets: USD, USDC, USDT, USD1, U, and BNB. The Funding Account will be officially renamed the Stocks Account in January 2027, with the exact date to be confirmed. Post-renaming, users may still use the aforementioned six tokens to purchase stocks via the Stocks Account, while deposits and withdrawals of other crypto tokens will only be processed through the Spot Account. A "One-Click Migration" button will be added to the Funding Account, enabling users to manually transfer crypto assets to the Spot Account. Assets not migrated by users will be automatically transferred in batches by Binance’s system starting January 2027, requiring no manual action from users, and their total asset balance will remain unaffected.

According to on-chain analyst Ai Yi (@ai_9684xtpa), the new address 0xf67…40F6b withdrew 1 million UNI from Coinbase two hours ago, worth $10.07 million at a withdrawal price of $10.07.

Per TradingBeats’ monitoring, the CXMT contract on Hyperliquid rose to $9.0063 earlier this morning, a roughly 14.5% rebound from its September 2 low, and just ~1.7% shy of its all-time high of $9.1654 since launch. Meanwhile, CXMT’s funding rate has turned negative again, currently at ~-0.0269% per hour, translating to a simple daily rate of ~-0.646%, meaning short traders will once again pay higher holding fees to long positions. The largest CXMT short holder (address: 0xf292) holds approximately $25.259 million in short positions, with an average entry price of ~$6.505, posting an unrealized loss of ~$6.391 million. This address has cumulatively paid a net ~$5.213 million in funding fees, with actual payments totaling ~$46,000 over the past 24 hours. If the current funding rate holds steady, this address will need to pay roughly $6,825 per hour, or ~$164,000 per day.

Beating AI Express (Dongcha News) reports that Ant Group's Bailing (inclusionAI) has open-sourced the Ming-Image-0.1-Design series, which includes two 6B models. The Design model is specialized in generating UIs, dashboards, infographics, and posters, while the Layer model splits a generated design into movable, replaceable, and editable transparent layers. Both models are released under the MIT license. The Design model supports structured prompts up to 8K tokens, enabling full-page generation in a single pass or direct creation of RGBA assets with transparent backgrounds. The Layer model can split a design into 2 to 9 layers at maximum, covering text, cards, main elements, and backgrounds, before reassembling them into the original design. Ant Group also simultaneously open-sourced Design Skill and PPT Skill: the former integrates design generation, asset decomposition, and front-end coding, while the latter converts a flat image into an editable PowerPoint file that allows modifications to text, images, and layout. In the open-weight sub-ranking of Artificial Analysis UI/UX Design released by Ant Group, Ming-Image-0.1-Design ranks first with an Elo score of 1082. The models are now available on OpenRouter and can be accessed for free.