R&D Meta-Strategy

A framework for thinking about automated R&D in 2026:

1. Try everything. The cost of development is going down very quickly. You can delegate so much more to agents. Try more things out, have lots of ideas and execute on them quickly.
2. Evaluate everything. Fail fast. Put your ideas in contact with reality. If this is hard or expensive, find ways to approximate the ground truth as early and often as possible.
3. Iterate. Use the information you attain. Tighten the feedback loops. Improve the system.