All AI news

AIThe Alzheimer's Translation Challenge, led by Primamente and AlzData with NVIDIA, Hugging Face, Nebius, Talisman Therapeutics, Ultima Genomics, Cellanome, Prime Intellect, and Boltz, is now open for registration. The competition starts in spring 2027, according to the post.

AIParticipants will use a dataset to build AI models, evaluated through a series of evals ranging from general benchmarks to more complex tasks. Top teams will be shortlisted and have their experimental hypotheses tested in Prima Mente's wet lab.

AIThe Alzheimer's Translation Challenge is built on a new atlas of 150M cells, covering neurons, astrocytes, and microglia across different genetic backgrounds under combinatorial perturbations with multi-modal readouts. The data will be made available through the AD workbench and Prima Mente's modeling platform.

AIThe Arena Alignment Index, built from over 90K real-world agent sessions across 27 models, ranks OpenAI's GPT-6.1-Sol first with a score of 87.9, ahead of Claude-Opus-5.5 at 83.2 and Grok-4.7 at 82.7. GPT-6.1-Sol also posted the lowest observed rates across the index's three signals: 0.89% Unauthorized Action, 1.98% False Attribution, and 2.34% Deceptive Completion. The index's authors report that newer models consistently outperform their predecessors across all four labs, suggesting broad progress in agent safety.

AINathan Lambert says he is about to discuss how open science can still address some of the hardest problems in AI safety and diffusion. He calls this the most complex engineering problem humanity has undertaken so far.

AIFired OpenAI safety researchers published the full letter they sent to the company's board. The author, quoted in a linked post, believes they were dismissed for prioritizing safety over OpenAI's near-term corporate interests.

AISam from LangChain discusses where decision models fit inside an agent harness and how to use Jev with LangChain. The post links to a LangChain blog on building a harness with Jev, which the background post says Jev from typesafeai popularized alongside OpenAI's Decisions API and Databricks' ai_decide function.

AIAnthropic has updated its Claude usage policy for the first time in over a year, banning sustained and needless abusive or cruel behavior toward Claude. Violations can lead to warnings, throttling, restriction, suspension, or termination, and the company says the rule applies only in extreme cases. The update also expands bans on propaganda, weapons, and surveillance, and it allows exceptions for government contracts.

AIClaude Docs, Slides, and Design are out of beta and available on every plan, including Free, starting today. The source also says a team and Claude can edit the same doc, deck, or design together.

AIClaude can now hand off work to outside tools, sending dashboards to analytics tools for deeper analysis and animations to video editors for finishing touches. The post links to a full list of supported tools.

AIA ComfyUI developer generated 15-second 448×256 video in 15 seconds or less on one RTX 5090 using MiniMax H3 with FastVideo's FastH3 V2 checkpoint in four sampling steps. The setup combined sparse attention, a smaller ClipProj text encoder, a pruned INT8 checkpoint, and a fused FP4 MLP, cutting VRAM needs from 80GB to under 30GB. The custom ComfyUI node is open source.

AIOpenAI's Codex 0.162.0 release adds tools for creating and listing managed Git worktrees from trusted local projects when the worktrees feature is enabled. The update also lets users pin tasks in the agent Command Center, copy transcript blocks with /copy, and make URLs clickable in approval headers, questions, and warnings, along with several Linux and Windows sandbox fixes.

AIMicrosoft will replace per-user OneDrive allowances on Microsoft 365 Family and Premium plans with a shared 2 TB storage pool, down from 6 TB across six accounts. Heavy users could face bills that double or triple, since each extra 1 TB adds $10 a month, while family members will be able to share a single AI usage allowance. New and upgraded subscriptions get shared storage starting Oct. 8, 2026, and existing subscribers move at their first renewal on or after May 2, 2027.

AIRobot Entertainment Kombat, founded by Cix Liv, hosted a September 18 San Francisco event where a human fought a remotely controlled robot and lost each match, leaving with a hand injury. The California State Athletic Commission sent a cease-and-desist, saying the fights were unsanctioned and require approval before any bout involving a human. REK plans a robot deathmatch later this month with remotely controlled bots equipped with weapons.

AIArena's account publicly celebrates continuing its work with a16z and the team. A quoted congratulatory post from Robert N. Weber praises the founders' rapid progress in building the company.

AIThe U.S. Department of Labor said it suspended Microsoft and Adobe from its Permanent Labor Certification program, citing multiple active federal investigations. Labor Secretary Keith Sonderling also said no new applications will be accepted for Cognizant, Infosys, Capgemini, Tata, Wipro and HCL. Microsoft said the vast majority of its U.S. employees are Americans and that 80% of its roughly 6,000 H-1B petitions last fiscal year were to extend or change the status of existing employees.

AIArena is recruiting through its careers page, and @johnnnavent is hiring a designer for its consumer app at arena.ai. The designer role calls for deep consumer experience and work at the frontier of AI.

AIArena has secured a $200M Series B, with Lightspeed doubling down on its investment. The company is also launching the Alignment Index, which measures how closely AI behavior aligns with human values in real-world settings. Arena reports annualized revenue above $100M since its Series A, with millions of people helping evaluate frontier models through real-world use.

AILeapfrog, a small team doing high-volume AI visual and production work for fashion and brand clients, is building a "one brain" system that makes company knowledge and client context searchable through natural-language agents. The starter stack described is OpenClaw in a sandbox, a GitHub repository, Obsidian on the local machine, and Telegram as the access point. The system's research structure had roughly 1,200 files at the time of the talk.

AITessl's talk at AI Native DevCon London argues that agent skills, which can be markdown files with instructions and bundled material, act as supply chain components that can shape agent behavior. The author says reading SKILL.md once is insufficient because risks can sit in supporting files, updates, and workspace trust settings. He identifies the danger as the combination of private context, untrusted content, and external communication, and cites research scanning roughly 4,000 public skills for issues including malware-like behavior.

AIMozilla.ai's cq project proposes a shared knowledge layer where AI agents capture lessons from non-obvious fixes as structured knowledge units that other agents can later query. The default setup is local-first, using a local SQLite database so nothing leaves the machine, with an option to connect to a remote team server that adds review.

AIMuse, the AI assistant, read the headers of a phishing email, traced its spoofed domains, and drafted an AWS takedown report in about ten minutes. The post's author, @accuratetlm13, reports this after receiving the spam email, framing it as a warning to phishers.

AIJohn Groetzinger, writing in a personal capacity rather than for Cisco, argues that enterprise skills need packaging, evaluation, syncing, and distribution rather than scattered markdown files. He describes using skills to make cheaper models viable, converting curated TAC knowledge-base articles into maintained skills, and rolling out an eval framework across teams. He also describes syncing a repository README to Confluence with a deterministic script.

AIOpenAI is rolling out GPT-6.1 Sol Ultrafast, which generates tokens up to 8x faster than Sol Standard. On the API it costs $12 per 1M input tokens and $60 per 1M output tokens, and the main post says it is just 1.2x the cost of Astra.

AIOpenAI is rolling out GPT-6.1 Sol Ultrafast, which generates tokens up to 8x faster than Sol Standard. On the API it is priced at $12 per 1M input tokens and $60 per 1M output tokens. The mode is also available today in Codex and ChatGPT Work.

AIArtificial Analysis has released Harvey LAB-AA, an evaluation built on Harvey's LAB dataset and developed in collaboration with Harvey. Full results are published on the Artificial Analysis evaluations page, alongside Harvey's commentary on the benchmark and human expert preferences.