"Like many others, I spent the last 2 months on a Claude Code bender, grappling with the fact that I no longer need to write code by hand anymore. I've been implementing AlphaGo from scratch (repo will be open sourced soon) to catch up on foundational deep learning techniques, and also to re-learn how to program with the full power of modern coding agents."
"For my "automated AlphaGo researcher" codebase, I created a Claude command /experiment which standardizes an "action" in the AlphaGo research environment as follows: Create a self-contained experiment folder with datetime prefix and descriptive slug. Write an experiment routine to a single-file python file and execute it. Intermediate artifacts and data are saved to data/ and figures/ subdirectories. All files are stored in easy-to-parse formats like CSV files that can be loaded with pandas."
Machines can now code and think effectively. A Claude-based coding agent implemented AlphaGo from scratch and automated research workflows to leverage modern coding agents' capabilities. The agent standardizes experiments via a /experiment command that creates datetime-prefixed folders, writes single-file Python routines, executes them, and saves artifacts and figures in easy-to-parse formats like CSV. Outcomes are summarized in report.md. Agents can run sequential experiments to optimize hyperparameters serially and can modify code itself, enabling an unconstrained search space that goes beyond traditional tuning systems like Gaussian Process bandits used by Vizier.
Read at Eric Jang
Unable to calculate read time
Collection
[
|
...
]