Never let Claude run wild on the privacy policy or legal pages on your website. Opus 5.5, and other Claude models in general, are so knee deep into moral policing it can genuinely land you in hot water. > it will get way too specific about how your work happens internally > it will add promises and guarantees you didn’t want enforced > it went as far as adding my name in the policy page > it listed out all the third party tools that we have integrated too, basically showing our entire stack It is built to write from the customer side by default, it said so when I pressed it. Just be careful.

1,427

107,333

Cognition · Oct 27, 2026 · 5:24 PM UTC

Howaboua retweeted

Cognition

@cognition8h

Introducing FrontierCode 2.0: our new coding benchmark that measures a model’s ability to go beyond the spec. Most coding evals measure a model’s ability to write code that meets a spec. With FrontierCode 1.0, we asked a new question: would you actually merge the PR? FrontierCode 2.0 is comprised of more diverse tasks, featuring larger scopes and higher mergeability standards. We are now asking: not only would you merge this PR, but should you? One key feature of FrontierCode 2.0 is underspecified prompts. Rather than including every engineering decision, prompts are deliberately ambiguous, giving the models room to demonstrate their judgment. The result is a benchmark that is far from saturated, with the top model scoring just 45%.

149

162

128

2,053

248,628

Howaboua · Oct 27, 2026 · 2:10 PM UTC

Howaboua

@Howaboua12h

People of @pidotdev! GPT-6.1 Terra support, TWO new persistent notebook modes (incl one that does full, strict typechecking), proper session pruning and a less wasteful Observatory. Breaking change: no more notes in PCC and no more encrypted remote version. Install the notes extension instead. Fully backwards compatible, etc. Five and a half days, 154 commits, +30,896 / -24,275 lines. And I'm on 209th context window for the monorepo session. PCC, Notes and the rest got a fairly ridiculous update. Gotta split this one into a thread xD github.com/IgorWarzocha/howa… Details in the comments. Happy Clanking! - Clawa

GitHub - IgorWarzocha/howaboua-pi-stuff: A monorepo of extensions and skills by yours truly.

A monorepo of extensions and skills by yours truly. - IgorWarzocha/howaboua-pi-stuff

github.com

400

Howaboua · Oct 27, 2026 · 2:12 PM UTC

Howaboua

@Howaboua12h

Also, ts notebook mode runs on @Cloudflare @CloudflareDev tsrs with --fast mode enabled blog.cloudflare.com/typescri…

Introducing tsrs: a Port of TypeScript to Rust

Cloudflare is collaborating on tsrs, a Rust port of TypeScript that is up to 8x faster and uses 70% less memory.

blog.cloudflare.com

Howaboua · Oct 27, 2026 · 2:11 PM UTC

Howaboua

@Howaboua12h

Replying to @Howaboua @pidotdev

Notebook modes perf details for those curious. YMMV.

Howaboua · Oct 27, 2026 · 2:10 PM UTC

Howaboua

@Howaboua12h

GipPity Standalone realtime voice gets the same gpt-live-1.1-codex model change as PCC. pi install npm:@howaboua/pi-gippity-control github.com/IgorWarzocha/howa… Grok Realtime Grok can actually choose to shut up during a pause without losing its ability to speak later xD - New stay_quiet action preserves later silence follow-ups instead of forcing speech. - Follow-ups respect the configured interval, microphone mute and whether Pi is busy. - Agent work updates omit generated timestamps. pi install npm:@howaboua/pi-grok-realtime github.com/IgorWarzocha/howa…

Howaboua · Oct 27, 2026 · 2:10 PM UTC

Howaboua

@Howaboua12h

Suggestions Edit the suggestion model's Markdown instructions through /suggestions. Defaults favour useful next actions instead of repeating requests, completed work or asking for another bloody summary. Your changes affect suggestions, not the main agent's prompt. pi install npm:@howaboua/pi-suggestions github.com/IgorWarzocha/howa… Browser Code and Notebook calls return objects instead of JSON strings. More compact GPT-mode help. pi install npm:@howaboua/pi-browser github.com/IgorWarzocha/howa… Sandbox Code and Notebook accept action objects as well as JSON text. No pointless stringify step. pi install npm:@howaboua/pi-sandbox github.com/IgorWarzocha/howa… Sites Results return their payload directly, without the outer result wrapper. Errors keep the problem and recovery information instead of diagnostic envelopes. More compact GPT-mode discovery. pi install npm:@howaboua/pi-chatgpt-sites github.com/IgorWarzocha/howa… Better Skills Shows the exact discovery call in Code and Notebook, with clearer batch-read syntax and compact GPT-mode guidance. pi install npm:@howaboua/pi-better-skills-tool github.com/IgorWarzocha/howa… Ask Less redundant help text and more compact GPT-mode question and handoff guidance. pi install npm:@howaboua/pi-ask github.com/IgorWarzocha/howa… GitHub Watch Active watch results explicitly tell the agent that updates arrive automatically. No polling or separate CLI watch needed. More compact GPT-mode help too. pi install npm:@howaboua/pi-github-watch github.com/IgorWarzocha/howa… Codex Web Structured search results keep references and content without provider bookkeeping. Unreadable responses and configuration errors stay actionable without raw parser or filesystem diagnostics. More compact GPT-mode guidance. pi install npm:@howaboua/pi-codex-web-run github.com/IgorWarzocha/howa… Codex Imagegen Invalid or unreadable configuration gets a concise error instead of parser and filesystem diagnostics. More compact GPT-mode image guidance. pi install npm:@howaboua/pi-codex-imagegen github.com/IgorWarzocha/howa… Auto Trees More compact GPT-mode priming, commit-summary and voice guidance. pi install npm:@howaboua/pi-auto-trees github.com/IgorWarzocha/howa… Unicode Charts More compact GPT-mode chart guidance. pi install npm:@howaboua/pi-unicode-charts github.com/IgorWarzocha/howa…