AI Coding Weekly

Theo makes three models argue before a merge

Theo now blocks his own merges until Opus 5.5, Muse Wonder Full and GPT-6.1 Sol have argued it out. Elsewhere, a claim that Claude-rewritten Bun has never had security issues got corrected by Bun's own author.

The big disagreement

Claim that Claude-rewritten Bun avoids security bugs gets corrected by Bun's author

Valerii Strilets posted that Bun, which he says was completely rewritten by Claude, has never been found to have these kinds of security issues. Jarred Sumner replied "Unfortunately, not accurate", and Theo said Bun got hit yesterday too, pointing at Bun's own note that v1.4.4 ships a fix plus one for a Brotli issue.

Jarred Sumner
@jarredsumner
X
Unfortunately, not accurate
Oct 23, 2026 · View on X

Read the full story

Receipts

Theo wrote a council skill that makes models argue before merging

Theo says he wrote a "council" skill that forces his agent to argue with Opus 5.5, Muse Wonder Full and GPT-6.1 Sol before it is allowed to merge. He has not posted what it caught or what it costs per merge.

Theo - t3.gg
@theo
X
I wrote a "council" skill that makes my agent argue with Opus 5.5, Muse Wonder Full and GPT-6.1 Sol before it is allowed to merge
Oct 23, 2026 · View on X

Read the full story

How they actually work

Pocock uses /grill-me one question at a time when the spec is fuzzy

Matt Pocock says he still reaches for "/grill-me about this, one question at a time" when he really has no idea what he is building and a single Q&A round could send things in massively different directions. He uses a separate /wayfinder prompt as an escape valve when a task will obviously take more than one session to plan.

Read the full story

Get the next one by email

Coding with AI, read daily so you do not have to. The experiments, the receipts and the arguments from engineers shipping real software. Not a changelog.