Thilak (@webtkdev) header
Thilak (@webtkdev) profile picture

Thilak

@webtkdev

Daily AI news. I build with it too.

 Chennai
webtk.dev/
Joined September 2010
  • Tweets2.7K
  • Following1.7K
  • Followers380
145Media
Photo by Thilak (@webtkdev)Photo by Thilak (@webtkdev)Photo by Thilak (@webtkdev)Photo by Thilak (@webtkdev)Photo by Thilak (@webtkdev)Photo by Thilak (@webtkdev)Photo by Thilak (@webtkdev)Photo by Thilak (@webtkdev)Photo by Thilak (@webtkdev)Photo by Thilak (@webtkdev)
Pinned Tweet
Thilak (@webtkdev) profile picture
Thilak
@webtkdev3h
Top 10 AI highlights from yesterday (Wed, Oct 28). Google's new #1 model is one most developers can't touch yet. GitHub now picks your models for you. And one robot showed how differently Claude and GPT write "hello world." Thread.
1
1
98
The new #1 on the Artificial Analysis index is a model you can't use yet. Gemini 4 Carbon, 59 points: - 88.2% SEC-bench Pro (finding vulns) - 71.0% CWE-bench (patching) - $2/$10 per M tokens at launch Cyber defenders first. Developers: "soon." What would you test first?
20
GitHub now picks your model, and its top tier lands within 2 points of Claude Code on Opus 5.5 Max. Copilot Fusion modes: - Single: one model - Cascade: cheap draft, judge escalates - Critique: one drafts, others review About half the cost per task. Would you let it pick?
16
10/ Artificial Analysis gave one robot the same task with two models. Claude Opus 5.5 dabbed out "hello world" dot by dot: 3,270 arm moves in 34 minutes. GPT-6 Astra wrote in continuous strokes: 514 moves in 13 minutes. teapawt.org/ArtificialAnlys/status…
Artificial Analysis (@ArtificialAnlys) profile pictureArtificial Analysis
@ArtificialAnlys8h
Same robot, same task, two models with very different writing styles ➤ On the left, Claude Opus 5.5 dabs out "hello world" dot by dot: 3,270 arm moves over 34 minutes ➤ On the right, GPT-6 Astra writes in continuous strokes: 514 moves in 13 minutes From AA-Robotics v0.1, seen from the model’s view See more examples here: artificialanalysis.ai/articl…
1
28
My take: yesterday was about the layer around the model. Routers pick it, reports price its work, specs let it touch lab hardware. The model is turning into a part, not the product. Which layer are you building on?
6
Copilot agents no longer need a person's login. GitHub's server-to-server auth (preview): - GitHub App, read-only Copilot Requests - 1-hour installation tokens - billed to the org, its budgets apply - per-app usage in CSV My take: overdue. Whose token runs your bots today?
16
Theo won't touch a product that needs a dashboard. My read: within a year, new dev tools launch with a CLI or MCP server first and the dashboard second. The agent is the user now. Which dashboard-only tool do you expect to lose users first? teapawt.org/theo/status/2115260769…
Theo - t3.gg (@theo) profile pictureTheo - t3.gg
@theoOct 28
If I have to go to a dashboard to use your product, I'm not going to use your product. If anything, I'm more likely to tell Claude to rebuild your product.
9
Top coding models, scored on "would a maintainer merge this?" FrontierCode 2.0 (1.1 in brackets): - Opus 5.5: 45.5% (54.6) - GPT-6 Astra: 42.8% (53.3) - Grok 4.7: 28.6% (47.6) - Gemini 4 Argon: 26.3% (47.5) - DeepSeek V4.1 Flash: 19.3% (43.9) - Kimi K3: 17.7% (44.2)
37
Your agent fixed the bug. In one place. Paste after every fix: "Search the repo for every other place this bug or pattern can happen. List each as file:line. Fix it, or tell me why it's safe." The same bug often hides in a sibling path. What does your agent miss most?
1
12
Infostealers now hunt your coding agent's login files. Aokigahara, a Windows stealer, grabs: - .claude/credentials.json - .codex/auth.json - Cursor's state.vscdb - .copilot/hosts.json - .aws/credentials My take: treat agent logins like SSH keys. When did you last rotate yours?
22
"Make it a LOT more fun" now works as a prompt. Simon Willison says Claude Opus 5.5 crossed a line for vibe coded games: - 4 browser games he enjoys - none over 4 hours of his work - he never looked at the code GPT-6 Astra isn't there yet, he says. What would you build first?
34
Load more