Should you RL your compound AI system or optimize its prompts? We think both! 🤯 A short preview of work co-led with Noah Ziems and Lakshya A Agrawal!👇

🔥

the wheel of tAIme

i have it now. no need to post the waiting meme.

🏋️ At Moxxie, we invest in grit and glutes! Thank you to the founders who joined us Barry's working on product market fit-ness 🙃

This is such a great point

Thank you OLZHAS for your contribution

This is a great point. It’s like Google saying “hey our employees drink a lot of water, so let’s go build our own water treatment plants”. Just a bad use of their time and resources.

This is a remarkable observation by Ankur Goyal. Models will improve faster than whatever decision-making engine you’ve built. So to take advantage of that, it’s better to build a system that pushes more decision-making to the model vs retaining it within your own engine. This way,

are you support vector machines old?

Good stuff

OpenAI achieves gold medal performance at the 2025 International Olympiad for Informatics. 6th overall in terms of score. 1st place among AI participants. It competed without internet or RAG.

🎯

this is the right interpretation of this meme

Incredible opportunity for anyone interested in joining this rocketship!

We're having so much fun Moxxie Ventures doing healthy founder events. This Thursday, we're doing a founder hike in SF followed by smoothies. Who wants to join? DM me!

Nice post on RLVR/RLHF libraries. Plus good comments in the thread

The four horsemen

💯

Good post on infra startups by Gokul Rajaram. Infra is increasingly becoming PLG driven!