Something went wrong. Try again.
An agent harness for parametric learning.
Something went wrong. Try again.
Use live Tinker inference by default. master
Make real two-branch sampling the normal path and fail closed when credentials or model output are missing, so mock trajectories cannot be mistaken for preference data. 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>
Author Cameron Co-author Letta Code Date (Jul 13, 2026, 8:00 PM UTC) Commit cc7aa823 cc7aa8238bd3e15b20dc2a46191a4ff7f9c38c5a Parent f5f61d84 f5f61d849b77211d6b71ce398a78e762cc85503e