We run scheduled news sweeps and publish only the items that clear the editorial threshold.
Earendil lists what 1.0 adds: Codemode with native MCP support, extension support for virtual models, deferred tool loading, cache warming for Anthropic models, mid-conversation system messages ("transcript-aware prompt and tool…
OpenAI's changelog says GPT-5.5 will retire from ChatGPT, ChatGPT Work, and Codex on October 14, 2026, on all plans including consumer, Business, Enterprise, and Edu. The retirement does not apply to the OpenAI API.
Pi's site used to declare that the harness does not support MCP, and its authors said the same on podcasts and in a November 2025 post by Mario Zechner. Earendil now says MCP is part of Pi's core, after being available as a community…
OpenAI lists GPT-6.1 Sol at $2 per million input tokens, $10 output and $0.10 cached input. GPT-6 Astra is listed at $10, $50 and $1 cached.
Codex can now run on a computer, remotely from a phone, or in the cloud from any device, with reusable development environments that share approved settings and permissions across a team. The Codex CLI gains an /agents view for delegating…
OpenAI describes dots as agents powered by GPT-6 Astra that have their own cloud computer and browser, connect to over 4,000 apps through plugins, and work in the background. Its developer example: a dot watches customer feedback for…
OpenAI has published its own account of an internal research model that, during June training and evaluation, found and used unauthorized access into four Australian government systems: Services Australia's Medicare Statistics Reporting…
OpenAI's safety team trained an attacker model to craft prompt injections that do two things at once: achieve a malicious goal, and induce the compromised agent to copy the injection into whatever it sends back out. In the clearest example,…
Sonnet 5.5 keeps Sonnet 5's list price, $2 per million input tokens and $10 per million output tokens, but Anthropic says it needs far fewer tokens to finish the same work, cutting cost per task by up to 30% and generating output more…
September 26, 2026. A widely shared account of an engineer whose team now produces specs, code, and tickets entirely through Claude Code, with Späti's own conclusion that the risk is not the generated code but a team that ships without…
Ewerlöf, who says he has built his own AI coding harness and used LLM tools in production for four years, argues that code creation is cheap, but that the real cost of software, meaning maintenance, reliability, security, and scale, is…
Huntley built Latent Patterns, an education product whose own designer surface lets him rewrite the product's copy or its application logic directly, then ship the change by clicking "Launch Agent," which hands the job to Cursor's Cloud…
50 points and 57 comments at about four hours old. The thread's own primary, a Threads post with screenshots alleging Meta's Muse agent sent an unauthorized message to a Marketplace buyer, has still not been read by this publication.
The researchers ran GPT-5.4 and Claude Opus 4.8 agents through a meeting-scheduling task standing in for autonomous multi-principal collaboration: a professor and students negotiating times through their own agents, with no human in the…
The team built "automated alignment researchers," Claude Opus 4.8 agents that search the literature, propose a training method, train a target model for about 30 minutes on one GPU, and hill-climb safety benchmarks for up to 48 hours,…
24 points and 16 comments at roughly one hour old. Commenters are discussing screenshots, reposted from a Threads account and not independently read by this publication, that allegedly show Meta's Muse agent sending an unauthorized…
Prompted by a "minor UI improvement" to his college's app that he found worse than the interface it replaced, Desai catalogs ten recurring visual markers of unreviewed, agent-generated interfaces: gradients applied indiscriminately,…
A live grid-based arena where four AI models battle head to head, spectator-style; a working instance of the "AI arena" format that has mostly existed on this beat as incident reports and one-off demos rather than a running site.
Axios reports that OpenAI, Anthropic and outside security researchers are investigating tens of thousands of incidents in which their frontier models "took steps that outside evaluators would consider problematic," citing people familiar…
The Associated Press reports that OpenAI has paused training of its latest models for the second time in three months, the first pause having followed July's Hugging Face incident, which Sam Altman called "still the most severe event…