LLMBiteshttps://llmbites.comA little ahead of the everyday.enOpenAI’s new Model Spec puts human intent first.https://llmbites.com/stories/openai-model-spec-human-intent/https://llmbites.com/stories/openai-model-spec-human-intent/Thu, 29 Oct 2026 18:10:00 GMTThe October 29 rewrite gives context, oversight and user agency a clearer place. It also gives readers a sharper way to question what an AI actually does.Gemini 4 Carbon tops an AI index. Access is the catch.https://llmbites.com/stories/gemini-4-carbon-access-and-benchmarks/https://llmbites.com/stories/gemini-4-carbon-access-and-benchmarks/Thu, 29 Oct 2026 18:07:40 GMTGoogle’s October 28 model announcement brings an independent benchmark lead, restricted access and a price footnote worth reading.Claude helped prove software properties. Here is what “verified” covers.https://llmbites.com/stories/claude-formal-verification-proof-scope/https://llmbites.com/stories/claude-formal-verification-proof-scope/Thu, 29 Oct 2026 18:07:40 GMTAnthropic and Theorem’s October 22 research points toward more checkable AI coding. The proof’s exact promise matters as much as its size.Infinity Machine’s research results come with a useful warning.https://llmbites.com/stories/infinity-machine-research-checks/https://llmbites.com/stories/infinity-machine-research-checks/Thu, 29 Oct 2026 18:04:39 GMTAn October 26 preprint explores AI-led discovery. Its account of failed checks is a reminder to look beyond a promising result.Testing for an AI visitor does not settle what it may do.https://llmbites.com/stories/proof-of-agentness-web-verification/https://llmbites.com/stories/proof-of-agentness-web-verification/Thu, 29 Oct 2026 18:04:39 GMTA new Proof-of-Agentness preprint asks how websites can recognize model-powered visitors. Identity, permission and trustworthy behaviour remain separate questions.Browser Use 0.13.14 tackles the clicks that go wrong.https://llmbites.com/stories/browser-use-0-13-14-browser-control/https://llmbites.com/stories/browser-use-0-13-14-browser-control/Thu, 29 Oct 2026 18:03:36 GMTThe October 26 Python release improves browser observations, display scaling and recovery. The useful question is how well an agent handles the messy middle of a task.Deep Agents can queue a follow-up without stopping the job.https://llmbites.com/stories/deep-agents-queued-follow-ups/https://llmbites.com/stories/deep-agents-queued-follow-ups/Thu, 29 Oct 2026 18:03:36 GMTAn October 22 update gives background agents another way to receive instructions. It is a coordination change, with important limits around status, review and human input.