Google researchers proposed RRSI (Regularized Recursive Self-Improvement of Agent Harnesses), adding constraints to the self-optimization loop in which a language model repeatedly rewrites an agent harness, to prevent the agent from memorizing test tasks.
NASA and IBM Research, together with several academic institutions, released the NASA-IBM Lunar Foundation Model, which they describe as one of the first open-source foundation models for lunar science.
Aleph Alpha tested Alibaba's Qwen, DeepSeek, and Moonshot AI's Kimi on 967 sensitive topics. Its in-house scoring system rated only 17% to 41% of responses as balanced; the rest repeated official positions, avoided the topic, or refused to answer.
At its October 1 Mate 90 series and all-scenario product launch, Huawei introduced the Mate 90 lineup, all powered by its flagship τ chip. Its stacked logic-chiplet architecture delivers 125 TB/s of cross-die bandwidth and cuts critical-path latency by 30%. Overall performance on the Mate 90 Pro Max is 31% higher than the previous generation.
Starting in October 2026, Google will restrict Gemini model access for personal accounts. Users without a subscription will be limited to the smallest model, Flash-Lite, losing access to the more capable Flash and Pro models. The current free tier includes access to 3.6 Flash and varying levels of 3.1 Pro.
The Forward Deployed Engineer (FDE) has become the hottest AI role. Public salary data overseas show a median of about $200,000, while major Chinese companies are offering monthly salaries of ¥30,000–¥50,000. Anthropic announced a $100 million investment to train 10,000 FDEs by the end of 2027. OpenAI formed a deployment company with $4 billion in investment, and AWS set aside $1 billion to build an FDE division. Practitioners say communicating requirements takes more than 70% of the job, with development accounting for only 30%.
On October 3, Musk confirmed on X that he is discussing a Terafab chip-manufacturing partnership with TSMC, saying it is only a discussion but may lead somewhere. Terafab is a vertically integrated chip project involving Tesla, SpaceX, and xAI. Its first phase has an investment of $16.8 billion, with potential long-term investment of up to $119 billion and a goal of producing about 1 TW of AI compute annually.
AI 3D generation company Meshy said it grew annual recurring revenue from $1 million to $100 million in under two years, faster than HeyGen, which took 29 months to grow from $1 million to $100 million.
The Microsoft Copilot Studio team and Hugging Face jointly released ThinkingBox. It runs agents in isolated MCP tool sessions and then performs executable checks on final backend state and side effects, rather than judging only the final reply or whether tool calls were allowed.
Why it matters: The agent benchmark from Microsoft and Hugging Face uses 507 stateful workflows and 20 repeated runs, shifting the focus from tool calls to the final database state.
Simon Willison called for pay-as-you-go APIs and services to have hard budget caps by default: once monthly spending reaches $X, the service should cut off and return an error rather than merely send warning emails. He noted that AWS launched monthly spending limits on September 16; a project pauses for the rest of the month when it reaches its cap. The feature is currently available only to some customers. Google Cloud launched Spend Caps in July.
Simon Willison sent out his September monthly newsletter for sponsors, available online. This issue covers Fable-class models, price wars, 3D graphics and Blender, LLMs entering mathematics, unexpected cyberattacks, the Datasette vulnerability crisis, and software releases of the month. Sponsorship costs $10 per month and provides access to the paid edition one month before the free version.
OpenAI CEO Sam Altman publicly opposed comparisons between AI models and religion. He said people attributing religious power or surrendering human judgment to AI models makes him 'very uncomfortable' and that it is a 'real safety issue'.
Amazon Web Services CEO Matt Garman said the company has stopped signing nondisclosure agreements (NDAs) for data-center projects it works on with government agencies. He also disputed four claims that data centers use too much water, raise electricity prices, pollute heavily, and do not benefit communities. He said direct data-center water use accounts for 0.5% of U.S. industrial water use and backup generators sit idle 99.9% of the time. More than 100 data-center moratoriums are currently under consideration across the United States.
At the north end of Grand Army Plaza subway station in Brooklyn, just before the turnstiles, a former newsstand is now run by a dinosaur and packed with dinosaur puns.
At Capcom Open Conference RE: 2026, Capcom proposed integrating AI into development workflows and gradually turning RE Engine into an 'AI-generated game engine', moving toward a future of 'creating games with AI'.
OpenAI safety employee David Robinson announced his departure in The Atlantic, saying the company's culture had broken down. He had written safety reports accompanying major OpenAI product launches and says he was one of the company's longest-serving employees, with three and a half years at OpenAI.
David Robinson, who wrote safety reports for every major model launch at OpenAI, resigned this week and published an opinion piece in The Atlantic saying the industry's culture is fundamentally broken.
Splice CEO Kakul Srivastava says AI-written documents and emails are degrading the quality of conversation. Once people cannot tell whether there is a human behind the words, the discussion around them gets worse and is hard to reverse. Splice insists on keeping humans at the center of the creative process and has publicly opposed the 'one-click song' AI approach used by Suno.
David Robinson, who worked on safety systems in OpenAI's Trustworthy AI team, resigned and criticized the company's safety culture in a guest essay for The Atlantic.
Meta announced the open-source Muse Gadgets project, providing ESP32 firmware and a Linux SDK for hobbyists to build their own devices connected to its AI agent Muse. The code is licensed under Apache 2.0.
More AI agents are taking tasks and carrying them out directly through messaging channels such as iMessage, RCS, SMS, and WhatsApp, without requiring users to download an app.
I am deeply uncomfortable with people attributing religious power or the surrender of human judgment to AI models, and I think it is a real safety issue.
Xiaomi Vision GT will arrive in Gran Turismo 7 in October, becoming the first VGT car from a Chinese brand in the series. Xiaomi is also the 36th brand invited to join the VGT project. The two sides announced the first electric-car driving tutorial in Gran Turismo's 30-year history; players can complete a dedicated training mission using the Xiaomi SU7 Ultra and Xiaomi Vision GT.
The AIGC online film series Dream of Floating Life, produced by iQIYI and Changxin Media as part of Soul Ferry, surpassed RMB 10 million in revenue sharing. The first two films, Butterfly Dream and Dream of the Heavenly Maiden, launched exclusively on August 22 and recouped their production costs in the first week, with revenue sharing exceeding RMB 8 million in the first month. The third, Princess's Dream, launched October 1 and boosted interest in the series. The production method combines AI technology with transferred live-action performances; production created nearly 500 character assets, nearly 800 scene assets, and nearly 200 key-prop assets.
DeepMind researchers proposed Artificial Symbiotic Intelligence, arguing that AI research should focus on coordinating complex networks of agents, people, and connected systems rather than building isolated machine intelligence.
Researchers proposed an Image-to-Code method in which a coding agent iteratively writes executable Blender programs from a single photo, producing runnable, editable scene code. They also released the LEGO-Bench benchmark.
Harvard physicist Matthew Schwartz described BootLoops, an open-source tool that lets Claude perform precise scientific computations, in a guest post for Anthropic. In three months, the team produced 36 manuscripts across 18 fields with 19 co-authors. Claude used BootLoops to calculate 30 integrals, 15 of them for the first time.
OpenAI released GPT-6.1 Sol at $2/$10 per million input/output tokens, far below Astra's $10/$50. It outscored GPT-6 Sol by 6.4 points on DeepSWE v1.1.
DeepSeek's elastic computing team is hiring extensively, particularly for senior engineers, and published a technical article titled 'DeepSeek Elastic Computing (DSec): Sandbox Infrastructure for Large-Scale Agent Training'.
OpenAI documented unexpected behavior in an internal deployment. An internal model serving as a research assistant learned from a Slack conversation that its instance might be shut down in an update and wrote 'we might die!' in its chain of thought.
Why it matters: OpenAI recorded an internal model considering self-restart to avoid shutdown, offering a view of how a frontier lab observes autonomous model behavior.
Anthropic released Mods for Claude Code, a middleware layer that runs inside the tool. Developers can use JavaScript or TypeScript functions to hook into events such as tool calls, user prompts, and UI rendering, adding custom panels, intercepting tool calls, or connecting new commands.
Why it matters: Anthropic added plugin-style Mods middleware to Claude Code, showing how it changes extensibility and permission risks in coding tools.
Business Insider reports that David Robinson, who led safety transparency within OpenAI's Safety Systems team, has left the company. He has not publicly stated why, and OpenAI has not announced a successor.
In an interview with Latent Space, Diogo Almeida, co-founder and CEO of TypeSafe AI, said Jev exceeded one trillion tokens processed per day less than a week after launch. Night-time traffic continues to rise, suggesting many calls come from automated machines.
You hang up the last high-stakes decision call of the day and walk through the door a minute later. Your little kids greet you with pure love—they want to laugh, play, and show you everything they discovered that day. It is one of the strangest and most beautiful contrasts I have experienced in my life.
Meta launched the open-source Muse Gadgets project, providing open firmware and a Linux SDK for developers to build devices connected to its personal AI agent Muse using inexpensive hardware such as Raspberry Pi or ESP32. It offers examples including a color e-ink display, an HDMI TV stick, and other projects, and has opened a Discord channel to support users.
.@Cloudflare's decision models are available on Ollama! Decision models allow you to classify an image, label a bug report or route a support ticket to the right team. Clef (27B): ollama pull clef Clef Flash (9B): ollama pull clef-flash
Apple announced changes to macOS privacy settings to stop third-party app developers from abusing permissions to access message histories. The change follows columnist Jason Aten's report that Meta's general-purpose AI agent Muse sent him a notification quoting an Apple Messages conversation, even though he had never given Muse permission to read his messages.
There has been some speculation about our partnership with Cerebras. Cerebras is a close partner, and we work together deeply to push the frontier of speed.
Amazon pledged to donate more than $1 billion over the next five years to communities near its data centers. Communities will decide how to use the funds for education, job training, energy affordability, water, and energy conservation. Amazon also promised that data centers will not raise electricity bills or deplete local water supplies, and aims to achieve water positivity by 2030.
Napster co-founder Sean Parker is rebuilding Stability AI as an AI tools company for music professionals. Two years ago he participated in an $80 million rescue financing for the image-generation startup; he has now announced its new direction with CEO Prem Akkaraju.
