Context Drift
Evals
+9
Jun 9, 2026
•
21 min read
20 Lessons I've Learned while being humbled by RL (Field notes for fellow Data/PM folks making the leap)
What a handful of experiments has taught me about the pre-training bones of RL environments, rewards, verifiers, holdouts, and harnesses. I’ll share a future Part 2 that gets into optimizer rollouts and policy learning, once that’s done, kicking my a$$.
