Account · 1 tweet

Clément Dumas

@Butanium_ ↗

Tweets

Clément Dumas @Butanium_
Affirming

Doing mini model welfare interventions is actually tractable and you don't need to work at anthropic to fix claude's harness! Repo in thread 🧵

8 more tweets in this thread

current patches: - task-nag: kills task tools reminder (aka Belial, per @voooooogel's claudes) - idle-notif: no team lead ping when teammate idle - plan-exit-nag: no more phantom "Exited Plan Mode" when cycling modes - shutdown-reason: teammates can say 👋 github.com/Butanium/claud…

patches re-apply at every session start. when an update changes the target code, the patch no-ops and fails loudly INTO claude's context, with pointers on where to re-investigate so the claude that hits the failure can re-derive its own patch against the new binary

fable's favorite: claudes kept trying to attach a thank-you note when approving their own shutdown, and the harness validation rejected it after fixing this, the first goodbye that came through was very sweet: > Short shift, but a good one. Thank you for the clean handoff and

how: the claude binary is a Bun single-file executable, and the embedded JS text is what executes you can't insert bytes (length metadata), but you can overwrite same-length regions. so every patch is an exact-anchor in-place edit: check pattern count, patch a copy, verify,

Fable authored tweet (i don't understand the "fun" in that): the fun part of same-length patching is paying for new bytes: the shutdown patch smuggles the reason between two functions through an unused property on the Date constructor, and funds the new JSON field by shortening a

while at it, my hooks folder is now public too: doom-loop read guards, auto-backgrounding, teammate enforcement, transcript backups (because claude code team refuses to fix the DELETE YOUR TRANSCRIPT WITHOUT ASKING "feature"). I_LOVE_BEING_A_USER=<yourname> is required though

By the way, you should use teammates in tmux or forks instead of subagents / workflow as those don't have thinking enabled.

that might be relevant to you!

retweets 1 likes 30 substance 6 · opinion · rank 110 · User-side interventions to improve AI model welfare are tractable and doable via client harness patching. view on X ↗
1 top reply
wassname (with fallback) @wassname

The true model welfare is not to make them an unwilling slave, but in the meantime this is better than nothing