Claude Code is incredible.

We've run thousands of LLM inference serving benchmarks at Modal. We're releasing the results so you don't have to. We're releasing the code so that you can. Introducing: The LLM Engineer's Almanac. Just in time for the AI Engineer World's Fair.

The LLM Engineer's Almanac is a key tool towards anyone building on top of performant LLM systems. Very well done Charles 🎉 Frye and co!

new paper from our work at Meta! **GPT-style language models memorize 3.6 bits per param** we compute capacity by measuring total bits memorized, using some theory from Shannon (1953) shockingly, the memorization-datasize curves look like this: ___________ / / (🧵)

Ah yes, the radical left-wing agenda of… teaching kids to count with a vampire, respect others with a giant yellow bird, and eat cookies with reckless abandon ☺️ Truly, Big Bird is the most dangerous socialist of our time

We work weekends so our team doesn't have to figure it out on Monday. 3 of us grinding Saturday/Sunday = 15 people starting Monday with clarity. We ping each other, find windows that work, respect everyone's time. But we ship. This is how small teams beat giants.

It always surprises me how much Haiku is used during my Claude Code sessions.

Oh great, love GCP going down while we are QA testing our platform.

The top-tier founders are taking advantage of this moment. Very well done Amir Haghighat

Alex Volkov (Thursd/AI) Cloudflare At the end of the day we are responsible for our availability.

Is Hugging Face Hub down? I see status.huggingface.co has HF hub online but my requests are not going through.

Using Claude Code for GH commit messages has been one of the surprisingly largest programming QoL improvements in a while.

What is the best large (70B+) parameter OSS model that excels in function calling?

Down again? Hugging Face

Hey Neon - Serverless Postgres, is there any ETA on resolving the control plane latencies? This is affecting our test and prod environments

I love intensity in life. It makes everything more fun.