Diogo Almeida

just learned the origin of the name claude 😬

classic allbirds pivot p.s. did they give up?

if A, then they should crack down on the Goodhart-ed objective and remove it entirely if B, they will instead crack down on it by using the Goodharting as an excuse to add checks like query uniqueness that mostly optimize for distillation, but might have handwavy benefit to real coding

Conspiracy theory - what seems more likely: A. meta doesn't know that you get what you optimize for B. they want this to happen as an excuse to get a ton of usage from anthropic that looks real and isn't an organized distillation effort IMO B seems way more likely (especially considering the pressure to give the impression they have a SOTA model)

I'm going to be at founders you should know (newsletter.foundersysk.com/p/your-showcas…), to reveal a tiny bit more to the world about what we do 🤐🤏 You could sign up to watch, or just assume it will be 🔥 and apply before everyone else (typesafe.ai/careers)

@andrewdfeldman Chart scaling 🧐

Just a little bit faster than everyone else on @OpenAI GPT-OSS-120B. > 10 times faster than the leading @nvidia GPU provider

@s3nhxx bias correction

AI/LLM benchmarks in 2025 be like

it's funny how many people wrote up huge predictions for MCP without even looking into how LLM performance degrades when you add even 10 tools to them

Bad news for the "current LLMs are reliable enough, we just need the right wrapper" crowd: customerexperiencedive.com/news/klarna-re…

congrats @OpenAI on the SearchGPT launch

@swannodette @RobStuttaford @nikitonsky It also had very prominent benchmarks which made a big difference in React.js adoption IIRC.

@ArgesRic @swannodette Finally decided to join! Thanks for showing @swannodette the thread! He had great input!

@swannodette (OP here) I have to say that Om has changed quite a bit since we've left! ref-cursors would've solved our biggest issue with Om