Debrieft

Read online 80% faster with 1 click. Simple.

[Debrieft @thedebrieft]

"[...] Summarization has been absolutely huge for me, much more than I thought it would be." – sama for Forbes

~80% reading time saved with just 1 click ⤵️
https://t.co/1MN26DGMTs

[Debrieft @thedebrieft]

@dharmesh @karpathy This is kind of what Glean does.

[Debrieft @thedebrieft]

@cremieuxrecueil Sunscreen paradox

[Debrieft @thedebrieft]

@bcherny Restoring sessions from GUI? Good search for sessions?

[Debrieft @thedebrieft]

@levelsio @sonofatailor Love the hi-necks from them.

[Debrieft @thedebrieft]

@austin_rief @oliverbrocato

[Debrieft @thedebrieft]

@RazaRafiq1990 @_0xaryan We have a very simple `consult` skill written, that cross checks ideas/solutions with both, Gemini and Codex.

[Debrieft @thedebrieft]

@piotrek_nowy Finally.

[Debrieft @thedebrieft]

@levelsio When run with --debug, are you getting 529s too? Freaking annoying. Switched to Codex for that reason 😬

[Debrieft @thedebrieft]

@ClaudeCodeLog When 529 error fixes? It's unusable. Paying for Max and I had to switchi to codex.

[Debrieft @thedebrieft]

@RickRubin YouTube? :)

[Debrieft @thedebrieft]

@bcherny What about recent 529s? Paying for the Max plan and executing e.g. simple Bash commands takes tens of seconds.

[Debrieft @thedebrieft]

@NatalieMarleny @UltraLinx We will have this soon :)

[Debrieft @thedebrieft]

@AnthropicAI Is it possible (within Claude Code session), to spawn a subagent from my current conversation, and actively/manually work within it, to later come back to the "root" agent?

[Debrieft @thedebrieft]

Didn't know that Claude Code allows you to customize the progress/status bar.

[Debrieft @thedebrieft]

@sundeep It indeed is!

[Debrieft @thedebrieft]

@nomadmum81 @kosik_md Wait till they discover OpenEvidence :)

[Debrieft @thedebrieft]

@PtrPomorski True. Same case with „The Godfather of AI” saying in 2016 to stop training radiologists 🤷‍♂️

Gemini 2.5 is really good though. Plus Google is really stable, compared to e.g. OAI.

[Debrieft @thedebrieft]

@MSFTResearch When YouTube?

[Debrieft @thedebrieft]

@zakkohane Interesting. "LLMs occasionally hallucinated unsupported details, while physicians were more prone to omissions or cognitive biases [...]" – i.e., HITL approach still the only safe option. I wonder what healthcare's "self-driving automation levels" are, and where we're at.

[Debrieft @thedebrieft]

@gregisenberg True, but this has always been this way.