Are you noticing more "opinion" pieces decrying BNPL for "normalizing debt" while overlooking that credit cards do the same, but cost you much more? With @Affirm you pay less (if any) in interest and no fees of any kind. So who benefits from these hit pieces? You? Or the banks trying to protect their credit card profits?

To push the open source frontier for RL + LLMs, we need scalable, modular environments with real-world complexity, beyond math benchmarks. Today, we’re releasing *benchmax*. An open-source framework to build, run, & scale useful RL envs for LLM fine-tuning, with integrations to verl & verifiers (more coming soon!).

So AI-IDEs that know how to download and incorporate assets (eg responding to a prompt "find a free image of a wizard and use that for the main character") create a whole new kind of vulnerability ready for malicious code injection, right?

Quick spin before a long flight, whereupon I plan to do some completely unserious vibe-coding, Garmin face attached. Coincidence?! 🤯😂

Yep, I’m that strong.

1/ Can codebase-specific RL push the frontier for code LLMs? At @cgftlabs, we helped a client RL-tune Qwen-2.5-7B on their internal codebase for unit test creation, with coverage-guided GRPO. The result? It beats o4-mini & o3. Here’s how it works (link to full blog in bio) 🧵

Comparing perceived attitudes of Oura and Whoop apps: Oura: wow you slept 5h, that’s like better than yesterday, yay!, so let’s try to make it a bit more today, we can do this!! Maybe take it a little easy though, knowhatimean? Whoop: your recovery is 38%.

ICYMI – our Founder and CEO @mlevchin appeared on @ABCNewsLive with @dianermacedo to discuss how Affirm responsibly extends access to credit. Check out the full interview:

We get the job done!

Seoul unexpectedly based.