AI Daily Dev — October 29, 2026

Anthropic's AI for Science Expands to 20 Foundations, Institutes, and Research Projects

Oct 28: Anthropic brings AI for Science from a researcher-grant program to 20 named partners, among them the Simons Foundation, Alfred P. Sloan Foundation, Schmidt Sciences, NSF, Wellcome Sanger Institute, EMBL, the Francis Crick Institute, and the Allen Institute.
Includes the Model Hardware Standard (MHS), a research-preview spec co-developed with HHMI Janelia so Claude agents can safely drive real lab instruments; Allen Institute is building agent teams to analyze its open datasets.
Research initiatives covered: FlyWire (Princeton connectomics), LEAP (Columbia climate), AstroPath (JHU cancer imaging), gnomAD (Broad; 800K+ genomes), and the Tau Ceti open Lean math library.
Lands the day after Anthropic's drug-discovery and GSA announcements — a third pillar in a week-long push to position Claude as default science infrastructure ahead of the November IPO.

research anthropic.com

JetBrains InnovationHub's open-source agentic IDE lands at the top of GitHub Trending today with +1,068 stars (5.3K total, 273 forks) and 604 commits on main.
Each chat runs in its own git worktree and branch, so multiple agents can edit the repo in parallel without touching your working checkout.
Apache-2.0; built on the open-source pi coding agent rather than JetBrains' own Junie, with a SPEC.md spec graph, MCP server + Agent Skills support, and either a desktop app or a browser UI from the `thinkrail` CLI.
No in-product billing — model usage runs through JetBrains AI or your own provider keys.

tools github.com

Oct 27 Stanford Daily reports a large share of Stanford professors' trade books turned up in the LibGen trove Anthropic admitted using to train Claude.
HN front page at 437–465 points and 251–263 comments; thread focus is less on the $1.5B settlement than on how many academic authors land under the $3,000-per-book payout.
The July 20 class action was certified around 7M+ pirated books, with Judge Alsup separating training (fair use) from Anthropic's 'central library' (not fair use).
Lands as Anthropic's S-1 week brings the IPO narrative into contact with the training-data record for the first time at scale.

industry stanforddaily.com

Nov 2026 print feature interviewing founder Ilyas Lebleu on the volunteer group running sweeps for LLM-generated articles across Wikipedia.
HN front page at 437–449 points and ~200 comments — the editor community discussing fabricated sources, pattern-matched giveaways, and the project's public LLM-tell guide (later reused by the 'Humanizer' Claude Code plugin).
Context: Wikipedia adopted WP:LLM in March 2026 to prohibit LLM-generated or rewritten article content, with carve-outs for copyediting and machine translation.
Project framing is 'verify, don't ban' — AI output stays if it clears the same sourcing bar as any other edit.

community spectrum.ieee.org

cloudflare/security-audit-skill is #1 on GitHub Trending this week with +16,002 stars (51.9k total), the biggest weekly velocity of any agent-skills repo.
A six-phase coding-agent skill — reconnaissance, mapping, discovery, exploitation attempts, verification, reporting — that spawns independent reviewer agents to re-check each finding before writing a `findings.json`.
MIT-licensed; one-line install for Claude Code, Codex, Cursor, Windsurf, or Cline: `npx skills add https://github.com/cloudflare/security-audit-skill --skill security-audit`.
Signals the agentic skills ecosystem is moving past design cleanup into security — the second wave after last week's Hallmark/impeccable/design-skills surge.

open-source github.com

Oct 28 drop from MIT-spin-out Liquid AI: a new foundation-model family positioned as domain-specialized rather than one-model-for-everything, covering text, vision, audio, and time-series in a single architecture.
First HN surface at 122 points / 50 comments — small ripple but the first LFM3 release after a year of LFM2.5 vision-language variants (the 3B shipped Aug 12).
Pitched for edge and specialized enterprise workloads; Liquid Foundation Models' core claim is state-space-plus-attention efficiency at parameter counts well below standard transformer peers.
Open-weight variants and benchmarks on the Liquid blog; no API pricing announced in the launch post.

models liquid.ai