Haha. Well, if you want to read Sheryl Hsu’s excellent last paper just before joining OpenAI, read LeReT at ICLR’25. She showed that using DSPy’s optimizers to diversify the prompts used for sampling trajectories improves RL on multi-step programs. Threads below.

TINAD is one of the few physical books that ironically I have lost

We are at the brief moment in LLM capabilities where for the next few years everything will become an IDE before we shift to higher leverage management of the same tasks

This isn't just -- it's

using claude code/ai assistants for organizational tasks is the new trendy thing clean your desktop before ai does it for you

New paper: Reflective Prompt Evolution Can Outperform GRPO. It's becoming clear that learning via natural-language reflection (aka prompt optimization) will long be a central learning paradigm for building AI systems. Great work by Lakshya A Agrawal and team on GEPA and SIMBA.

🗽NYC AI builders! Join us Thu Aug 7 for a co-working day hosted by Station3 - Crypto Hub NYC Connect with founders & engineers + Get live demos from local builders: AR robot repair multiplayer AI agents Camcorder AI ☕ Free coffee/snacks + jazz at 5:30pm 📍 Drop in 10am-5pm

AGI is when a compound ai system can solve a new destiny raid in 24 hours

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!👇

Incredibly grateful to the team!! isaac 🧩 Liheng Lai Chen Qian KaiqiangSong Meng Jiang Dan Klein Matei Zaharia Karel D’Oosterlinck Christopher Potts Omar Khattab

she has literally no idea they released gpt 5 today

THEY DID WHAT TO ME

Well, in the most lowkey way possible, and on a random Tue afternoon, DSPy 3.0 is out of beta. pip install -U dspy So many amazing people contributed to this. Thank you all! (Release notes below. Stay tuned for a ton of stuff over the next few weeks!!)

LETS GOOOO

~90% of non-technical people who use chatgpt that I ask call it Chat

the "use it or lose it" movement is going to have a big resurgence when people realize that they have lost it and become oneshot zombies rehab for people who have been one shot. who is building this spec ?

Is there an RL type follow up for the "Are Emergent Abilities of Large Language Models a Mirage?" from NeurIPS 2023?

Absolutely—you should drink another Spindrift and proudly add it to your can graveyard.

wearing a hoodie lets me lock in like I have horse blinders on

put the prompts in the bag lil bro