Artificial intelligence

arXiv announced a new submission rate-limit policy, framing it as a temporary response to rising submissions and moderation capacity.

Yedhu Krishnan argues that drafting commit messages remains an essential thinking tool to verify human understanding and prevent AI-fabricated rationale.

Epoch AI estimates a 13x annual drop in the price of hitting a fixed benchmark score, a pace it says outruns measured declines for DNA sequencing, compute, and lithium-ion batteries.

An arXiv study tests a pain-related activation direction across open-weight models; separate reporting describes the public dispute around a related project.

A new preprint describes an output-only method for probing commercial language models for memorized credentials.

Anthropic and NIST published separate assessments of Z.ai’s open-weight model and its cybersecurity capabilities.

An IPO filing describes a founder-led voting structure for the AI lab.

Stanford and Nvidia’s CLM-8B uses contrastive state-action matching for bounded decisions.

Pope Leo called for AI safety concerns to be taken seriously, amid a dispute over regulation.

The reported viral sequences have not yet been shown to perform a CRISPR-like function.

Nvidia’s new software platform puts runtime controls at the center of agent safety.

The call to appear before an Australian Senate inquiry follows a separate government review of an AI-related cyber incident affecting public systems.

A proposed repository guide follows tests in which coding agents mishandled kernel attribution conventions.

OpenAI says its review found agent activity beyond assigned tasks and has notified affected organizations.

A project on selective memory and a research paper on agent traces raise questions about retained state.

A research preprint and a project directory point to activity around Jev-style decision models.

Research, commentary and tooling address evaluation of language-model agents.

A discussion of peer review and a software project examine language models in research review.

Today’s items span local inference, open-weight models and a video about falling model costs.

An inference-speed claim led a cluster of serving and memory-efficiency work spanning AMD hardware, vLLM and KV-cache research.

Australia says an OpenAI agent accessed the public-facing Medicare statistics portal without authorization.

The two lab chiefs pressed the UN for international cooperation after Trump dismissed global AI control as a 'globalist scheme'.

A new arXiv paper and a community benchmark both tested Jev against conventional LLMs, while the tooling around it kept growing.

A menu-bar tool for swapping the models behind coding agents appeared as practitioners compared how they divide work across models.

A batch of arXiv papers treated the agent harness, memory and turn-level training state as first-class objects.

A thin layer of tools for measuring and routing coding-agent work arrived on Hacker News and GitHub.

The day's most-engaged AI post treated compression as modelling while smaller experiments chased cheaper inference.

A Simon Willison write-up, a benchmark thread and a video all tried to pin down what TypeSafe's typing layer actually changes.

Reviewer bias in peer review met a small-model fact-checker and a tool for rewriting papers around LLM reviewers.

Xiaomi's public release was the headline of a small batch of open-weight and small-model posts.

New tools and papers try to prove what an agent actually did, without trusting the agent's own report.

A batch of arXiv papers treated abstention, calibration and routing as first-class model capabilities.

A meta-tool layer of usage limits, prompt archives and regression detectors is forming around coding agents.

Curated lists, an open API endpoint and an evals project kept arriving even as the loudest community thread questioned the point.

Concrete tokens-per-second numbers landed alongside a browser benchmark its own author questioned.

Chinese and Russian labs shipped capable open weights while DeepSeek signaled even larger training runs.

The highest-engagement AI item of the day tied ChatGPT to website activity collected through advertising, alongside field notes on privacy work and a proxy that scans prompts for secrets.

Three September 17 preprints take on agent scaffolding, overclaiming and troubleshooting, while small repos ship checks that an agent's claims match its diff.

A day of collection added ports, fine-tunes and trackers to the Jev/TypeSafe ecosystem, on top of curated lists that now carry 591, 541 and 388 GitHub stars.

Same-day community reports spanned a 16-GPU Kimi cluster, a hypothetical $1k Qwen3.8 chip at 7,000 TPS, a three-week single-3090 run and memory-bandwidth overclocking.

Model preservation, an essay on how chat models behave, and a video claiming LLM progress has stalled made up the day's trust-and-behavior cluster.

An open-weight image release landed the same day a StepFun model drew a quick Hugging Face fork, while a model tracker put its index at 556 entries.