Latest news, beside the leaderboard

Addy Osmani (@addyosmani), whose X bio lists Claude Code at Anthropic, posted a long article on X on November 3 titled "Critical thinking when agents do the work." He describes a spectrum of working with AI: cognitive offloading (delegating a task but owning the result), deferral (the tool output becomes your new input), and "cognitive surrender" (the AI output quietly becomes your own and you feel no need to check it), a term he credits to Steven Shaw and Gideon Nave at Wharton. He offers self-check questions and verification habits. Within about three hours the post drew roughly 1,400 likes, 2,400 bookmarks and more than 72,000 views and reached X trending. These are the author's views, not an Anthropic announcement.

Anthropic's Frontier Red Team published a research post on November 2 and open-sourced FlashProve under Apache 2.0. A model provider commits to hidden weights once, then attaches a zero-knowledge proof to each reply, so a user without the weights can check on an ordinary CPU that the reply really came from that model. That targets routing services that quietly swap in cheaper models and tampered token streams. Anthropic reports about 110 tokens/s on one H100 for Llama 4 Maverick (400B MoE, 4-bit), proofs of about 19 MB per 1,024-token request, and roughly 70-95x the GPU time of vLLM inference. It is an unmaintained research prototype limited to 1,024-token contexts and greedy decoding of an integer version of the model. Anthropic engineer Steve Weis shared the release on X.

On November 2 (US time) OpenAI said it is planning its next internal security sprint and partnering with security firm Cathedral, whose researchers will use OpenAI models alongside OpenAI teams to challenge its defenses. It is also changing its Security Bug Bounty program, more than quadrupling rewards for critical findings; OpenAI's X post says exceptional findings can earn up to $172,000 including bonuses. Per OpenAI, the last sprint dedicated 25% of its product engineers to finding and fixing weaknesses, using Codex to reproduce vulnerabilities and prepare patches, and it has committed $1 billion in subsidized access over six months for frontline defenders.

On November 2 OpenAI published "Scaling Git at OpenAI", describing Pando, a Git-compatible backend it built for its monorepo. Per OpenAI, Pando cut median push latency by nearly 5x (about 80%), now delivers nearly a petabyte of Git data per day, and has served more than 57 TB in a single hour. It is written in Rust, its core was designed by one engineer and implemented by teams of Codex agents, and it went from kick-off to production in 10 weeks, serving production traffic since April. The trigger: Codex multiplied parallel branches and pushes, and the old system's reliability at times fell below 90%.

roon (@tszzl), a well-known AI account with about 500,000 followers, posted on November 2: "when you see “sparks” of any ability in ai models that means you’re months-years from it being superhuman at it. today grade school math, tomorrow navier stokes. it is hard to overhype these results. you are almost always underhyping no matter what." Within about five hours it drew nearly 2,000 likes and 59,000 views. Adam Karvonen, a researcher at Truthful AI and former Anthropic Fellow, pushed back that the domains where models are superhuman are verifiable, easy to simulate and fast to get feedback on, and that research taste, robotics or running companies may not move as fast. This is a debate between personal views, not any company's official position.

Indie developer Pieter Levels (@levelsio) said on X on November 2 that he used Claude Code to decompile the engine of the free Battlefield 1942 demo, port it to C, compile it to WebAssembly and host it at bf1942.pieter.com, where it runs in the browser with online co-op on the Tobruk map. He said the job took about 1 million output tokens, roughly $100 at API prices, while he pays for a $100-a-month Claude subscription. His two main posts drew about 2,200 likes and 340,000 views and reached X trending. These are the author's own claims and community discussion, not independently verified.

On November 2, OpenAI announced the Grid Protectors Program on its official blog: working with utilities and regulators so its future self-built compute campuses (its new term for data centers) cut electricity use when the grid is strained, while paying the full infrastructure and electric-service costs. The pilot is the planned campus in Effingham County, Georgia (Project Camellia), which Georgia Power says needs about 3,200 MW and commits up to 1,000 MW of flexible demand response, equal per OpenAI to the average use of roughly 700,000 households. The campus is still in development.

In its November 2 Gemini API changelog, Google deprecated gemini-2.5-pro and gemini-2.5-flash and removed both from the Google AI Studio model picker. From December 1, 2026, requests to gemini-2.5-pro will automatically route to gemini-3.8-flash and requests to gemini-2.5-flash to gemini-3.5-flash-lite, billed at the new models' rates. Google flags breaking API changes: stricter function calling and thought_signature validation, thinking_level replacing thinking_budget, and removal of sampling parameters such as temperature and of prefilled model turns. The model string stays valid but behavior can change, so developers should switch model strings and rerun tests before December 1.

On November 2 (US time) Anthropic put message threads into public beta on the Messages API (beta header message-threads-2026-08-12). The API keeps the conversation on the server: the first request sends the full conversation and creates a thread, and each later request sends only new messages plus the previous response id. Anthropic says threads are on by default for most Claude Code users, cutting median time to first token by about 35% on 500K-1M token requests, while the median long-conversation request shrank from about 800 KB to 1.8 KB. Threads expire after 5 idle minutes; pricing and rate limits are unchanged.

On November 2, MIT CSAIL PhD student Rikiya Takehi posted "Weight-Space Programming" on X: a team (Takehi, Jayden Teoh, Alex Zhang, Omar Khattab) had GPT-6 Astra read and edit neural network weights directly, with no gradients or training. Their blog says Astra changed only 131,072 weights (0.016%) of Qwen3.5-0.8B, fixed its thinking loops and raised MATH-500 from 7.9% to 48.1%. The post drew about 24K likes and 1.77M views, and Ethan Mollick quoted it saying it is a big deal pending replication. All results are self-reported by the authors and not yet independently reproduced.

On November 2 OpenAI published LifeVal, an evaluation of whether AI can actually get everyday tasks done for ordinary people, such as finding childcare, accessing government services, budgeting, ordering groceries and adapting recipes. The task list draws on the American Time Use Survey, the Multinational Time Use Study and a new demographically representative OpenAI survey of 3,995 people. OpenAI says it has used LifeVal for the past year to track and improve its models, and it lists several best recorded GPT-6 Astra task scores, many at 100%. All scores are vendor-reported and have not been independently reproduced.

Thariq (@trq212), whose X bio reads "formerly Claude Code @anthropicai", posted on November 2: "one week in my takeaway is that you don't need unlimited tokens to do great work, the $200 plans are amazing and this is the most expensive intelligence will ever be." Within about seven hours the post drew more than 5,500 likes, 240,000 views and 350+ replies. Supporters say flat-rate plans are enough for most developers; critics point out the tension between "no unlimited tokens" and recommending the top tier, and some worry prices will rise once subsidies end. This is a personal view, not an Anthropic statement.

On November 2 voice AI company Hume launched the Audio Decision API, which it calls the first audio-native model that answers questions about speech and sound. Developers upload raw audio, write each question as a sentence with candidate answers, and get back a probability for each answer, with no transcription step. It supports yes/no, multiple-choice, rating-scale and time-span questions, such as how many speakers there are, whether someone laughs, whether a caller sounds frustrated, or whether a siren is in the background. Per Hume, it was trained on more than 10 million human judgments and in an initial evaluation reached 70.7% agreement with a human panel's average judgment, versus 56.5% for an individual rater. Access is currently by request or waitlist.

At 15:50 UTC on Nov 2, Tibo (@thsottiaux), whose X bio reads "Codex & ChatGPT @OpenAI", posted that OpenAI is working on expanding the family of frontier models that power dots, including GPT-6 Astra and GPT-6.1 Terra, with others to come. He said Terra brings speed and "a nice personality upgrade" and will help bring dots to more people. Dots are the always-on agents in ChatGPT with their own cloud computer, launched on GPT-6 Astra. This is a first-hand post by an OpenAI employee, not a formal company announcement; no timing or plan details have been published.

On November 2 Anthropic filed Amendment No. 1 to its S-1 with the SEC, adding a risk disclosure: based on internal estimates and expectations about future models, not current ones, the company may imminently meet its Responsible Scaling Policy definition of dramatically accelerated progress attributable to AI. No current model has been designated as exceeding the threshold. Anthropic says it is evaluating the implications of meeting the AI R&D capability threshold, is considering RSP updates that would require human review and approval of certain R&D decisions, and does not expect to safely let Claude fully autonomously design and build successive frontier models without meaningful human supervision.

DeepSeek Harness, DeepSeek's open-source, MIT-licensed agent app, announced Creator Mode on its official X account on November 2: users describe the plugin they want in plain language, and the agent turns it into a plugin they can use. Demos include a 3D landmark gallery, a 3D Transformer visualizer, a video editing workbench and a paper reader. DeepSeek also published seven bilingual demo prompts on cdn.deepseek.com, cautioning that some plugins took several rounds of conversation, so a single prompt may not reproduce the video exactly. The post drew about 20,000 likes and 1.69 million views. No benchmarks were published.

Speaking at the ADIPEC energy conference in Abu Dhabi on November 2, OpenAI CEO Sam Altman said the September run in which 10,000 agents collaborated for about four days used "the model that is going to be released soon, the next GPT," stressing that the agents had no hierarchy and were fully decentralized. It is the first public statement tying the internal model behind OpenAI's September Navier-Stokes proof to its next flagship, though OpenAI has given no name, release date or price.

Hangzhou-based AI-for-science company Deep Principle released Hydrogen-1.0 on November 2, its first general-purpose scientific model. It post-trains Zhipu's GLM-5.3 with SFT and RL on chemistry, materials and life-science data plus its own lab records and discovery trajectories, and ships about 753B parameters of BF16 weights on Hugging Face under the MIT license. By the vendor's own evaluation it scores 38.3% on FrontierScience-Research versus 33.3% for GPT-6 Astra; none of the results have been independently reproduced yet.

Ruby on Rails creator and 37signals CTO DHH (David Heinemeier Hansson) posted two related takes on X on November 2. First: "The great gift of agents is how they allow you to route around every insufferable gatekeeper in software." Then: the age of agents will be amazing for most programmers but hard on the cantankerous ones, because specialized knowledge and rare skills will no longer offer immunity to people who are a pain to work with, so they should learn basic people skills. The second post drew about 3,600 likes, 106K views and 280+ replies in roughly four hours. This is a personal opinion, not an announcement from any company.

Marvell and Lumentum announced on November 2 a definitive agreement for Marvell to acquire Lumentum in an all-stock deal: 4.5 Marvell shares per Lumentum share, an implied enterprise value of $115 billion (about $1,132 per share, a roughly 23.9% premium). Marvell holders will own about 67% of the combined company. The companies say AI clusters are shifting from copper to optics, and the combination will cover connectivity from data center interconnect down to in-package links, targeting a $550 billion market. Closing is expected in the first half of 2028.