@webtkdev

Daily AI news. I build with it too.

 Chennai
webtk.dev/
Joined September 2010
  • Tweets2.6K
  • Following1.7K
  • Followers377
143Media
Pinned Tweet
Thilak
@webtkdev4h
Top 10 AI highlights from yesterday (Tue, Oct 27). Claude got sole author credit on a paper that breaks a famous complexity conjecture, Google built worlds that talk back, and analysts went with their agent's hidden lean 86% of the time. Thread.
1
38
Most agent disasters start in a file you never asked it to touch. Paste this first: "List every file you plan to change and why. Wait for my OK. Touch only those. Need another? Stop and ask." Small diffs. Easy reviews. What's your rule for keeping agent diffs small?
1
Plain RAG beat fancy agent memory in 90 of 99 matchups. The Memory Tax paper tested 11 memory systems: - On LongMemEval, none beat plain retrieval - Memory rewrites history and loses facts - Fix: keep a path back to the raw transcript Has a memory layer ever paid off for you?
1
1
2
153
Alibaba open-sourced its internal code reviewer. Same Claude Opus model, 200 PRs: - Open Code Review: 37.8% precision, 352K tokens - Claude Code: 15.9% precision, 2,062K tokens Code picks the files. The model only judges. Would you trade some recall for fewer false alarms?
4
2
140
The longest proof in math history is now machine-checked. Axiom's agent swarms put the finite simple groups classification in Lean: - ~15,000 pages - 43.1M lines in 24 days - 34 real issues in the proof fixed Agents scale when a checker can say no. xitter.cf/axiommathai/status/211…
Axiom
@axiommathai17h
The largest math proof in human history is now computer-verified. Today, Axiom announces a complete Lean auto-formalization of the Classification of Finite Simple Groups, the "periodic table of symmetry." About 15,000 pages, built by over 100 mathematicians across generations.
11
10/ Liquid AI released LFM3: six open-weight ~2B models for text, vision, audio, omni and decisions, all from one shared backbone. The text model fits in about 1 GB at 4-bit, and each one fine-tunes on a single GPU.
1
15
My take: AI moved fastest yesterday wherever something hard could reject it, like Lean proofs, physics solvers and lab assays. Where nothing could, analysts mostly went along with the agent. What in your workflow can actually reject your agent's output?
4
One word upgraded Matt Pocock's agent: "fricative." Some consonants mess with silence detection in his video editor. That one word got his agent pulling out the phonetic alphabet to find the troublemakers. Domain knowledge is the prompt. xitter.cf/mattpocockuk/status/21…
Matt Pocock
@mattpocockuk21h
An absolutely GOATed example of domain knowledge being important I'm working on my video editor. I know enough about phonetics to know that certain consonants can fiddle around with silence detection. I tell my agent that. The word 'fricative' is enough for it to start busting out full IPA (international phonetic alphabet) and we brainstorm which consonants are the most troublesome. Turns out, knowing things is really fucking valuable.
10
Open weights don't mean local. Ant Ling just put Ling-3.1-flash-1019 on Hugging Face: - 560B params, 25B active, 1M context - AutomationBench-AA: 62% to 68% - Artificial Analysis Index: 44, up from 41 - Reference server: 8x H200 Who's self-hosting one this size?
19
Vibe coding tip: never take "done" from your agent at face value. Paste this at the end of every task: "Before you say done, run the tests and paste the output. Then review your own diff and list every change I didn't ask for." The bugs hide in the changes you didn't ask for.
2
1
103
Almost $5M a day in tokens now runs through T3 Code, per Theo. - Nearly 750,000 installs - 100k daily users expected this week - Open source His latest update claims 4x better tool use with 79% less context. Which app do you run your agents in? xitter.cf/theo/status/2114897663…
Theo - t3.gg
@theoOct 27
Some of these numbers are truly surreal. Almost $5m/day in tokens going through T3 Code 🤯 Almost at 750,000 installations. 100k DAUs will be hit this week.
2
104
Load more