Just 5 days after Jev rose to prominence, Cua has compressed its decision-making model to 2.8MB.

Beating AI News Flash: Vercel Labs engineer Chris Tate has integrated Jev into json-render, launching experiments to enable AI to directly compose UI at runtime. Json-render pre-prepares buttons, cards, input fields, data, and actions. Unlike large language models (LLMs) that generate full JSON token-by-token, Jev only determines which components to use, their order, and placement before handing off to a standard UI renderer. The new interface defaults to selecting core components in one pass, with a second pass for layout when required. In a demo generating an identical train ticket interface, the default JSONL mode took 3.21 seconds from request to completion, while Jev clocked in at just 0.88 seconds. Future app interfaces may no longer need to be fully hardcoded. UI can become runtime-decisive, allowing the same app to dynamically assemble different pages for different users, tasks, and states. Jev’s latest demo has pushed full generation time to under 1 second, at least bringing this type of interaction closer to real-time. Runway’s Solaris, released at the end of last month, is pursuing a similar direction but with a more radical approach. Solaris uses a world model to generate entire interfaces frame-by-frame, directly bypassing traditional code and components; Jev, by contrast, still relies on pre-built components and actions written by developers, only dynamically deciding how to combine them. One aims to turn software into a real-time generated video world, while the other is more like giving software a face that can be rearranged at any time.

According to monitoring by TradingBeats, the unrealized profit on Ethereum long positions held in the address of crypto figure "Big Brother Machi" Huang Licheng has expanded to $3.65 million, with the position value reaching $88.81 million; his Bitcoin long positions carry an unrealized profit of $112,000. He recently added 9,000 HYPE long positions with 10x leverage, at an entry price of $92.21.

Beating AI Express News: Y Combinator (YC) S25 edge AI startup Cactus Compute has launched Needle 3, a model focused on tool calling and structured extraction rather than general chat. For instance, when a user says "Set the living room light to 30%", Needle 3 only needs to locate the corresponding function and input "living room" and "30%" into the parameters. It is similar to the recently popular Jev but takes a different approach: Jev directly scores candidate decisions, while Needle 3 still generates outputs token-by-token, with its outputs restricted to executable tool calls. Needle 3’s single set of weights can be pruned to 2 to 20 layers. The 2-layer version is approximately 9MB, and the full 20-layer version is about 35MB, allowing selection based on the computing power of different devices such as smartwatches, Raspberry Pi, and mobile phones. After being fine-tuned specifically for Android operation commands, the 4-layer, 29-million-parameter version achieved a self-test score of 62.5, outperforming DeepSeek V4 Flash’s score of 60.5.

Beating AI Express: NVIDIA CEO Jensen Huang recently told CBS News that he rejects predictions AI will cause human extinction before 2030, stating "2030 cannot be the end of the world" and the probability of AI ending the world by 2030 is "0%". Huang noted AI technology can be safely managed, adding the industry should not slow development over potential risks. "We should move forward as quickly as possible," he said, but stressed products must never be launched to market before they are ready or have security flaws. Earlier, Anthropic employee Jacob Coxon warned in his resignation that AI researchers "genuinely believe AI could kill all of us by the end of this decade", sparking a new round of discussions on AI safety risks. Huang has repeatedly addressed AI safety issues publicly lately, while opposing slowing AI development due to potential risks. Bloomberg reported that as CEO of NVIDIA, the world's top AI chip supplier, Huang has long opposed AI doomsday claims.

Coinbase CEO Brian Armstrong posted on X that The Wall Street Journal is preparing a report blaming Coinbase and him personally for the failure of the U.S. CLARITY Act. Armstrong noted that the WSJ’s prior coverage of the CLARITY Act was “clearly hostile” and echoed the views of banking lobbying groups. He explained that in January this year, he opposed a draft version of the CLARITY Act from advancing to a committee vote, as that draft had significant flaws in areas including DeFi, tokenization, CFTC regulatory authority, and stablecoin reward mechanisms—flaws that could harm the crypto industry. He added the bill also lacked sufficient support to pass at that time. Armstrong stated that afterward, Coinbase collaborated with multiple parties to push for revisions to the bill, and all four aforementioned issues were resolved in the version submitted to the committee roughly four months later. The final version of the CLARITY Act sent to the Senate is “good,” he said, and he strongly supports it. Armstrong emphasized that his opposition to the early draft does not mean he is against the CLARITY Act itself; rather, it was aimed at advancing a more comprehensive crypto regulatory framework, and he will continue to push for fair, clear rules for the crypto industry going forward.

Cronos Labs has launched a governance proposal on GitHub, proposing to allocate 100% of product revenue generated by Ult and Cronos Launch to open market repurchases of CRO, followed by token burning. The proposal calls for monthly on-chain repurchases and burns, with each transaction hash made public, to establish a direct "product revenue — CRO repurchase — supply reduction" mechanism. It notes that Ult launched on September 17, while Cronos Launch went live on September 15. The previously approved "New CRO Era" proposal originally planned to distribute product revenue across multiple areas: staking rewards, growth and user acquisition, repurchases and burns, as well as R&D and operations. This new proposal would eliminate those allocations, directing all revenue to CRO repurchases and burns instead, with operations, infrastructure, and growth expenses covered by existing funds. Separately, Cronos Labs intends to use its strategic reserve to support future Cronos POS staking rewards. As CRO’s inflationary emissions gradually decline per prior plans, the reserve will be used to supplement staking rewards to maintain current Cronos POS reward parameters, with staking methods, lock-up periods, and reward structures unchanged. The proposal is currently in the discussion phase and will later be submitted for on-chain governance voting. The voting period is 14 days, with a quorum requirement of 33.4% of staked CRO; approval requires more than 50% of non-abstention votes.