You don't run experiments anymore. Your autonomous AI research agent does that. Your job is to write program.md, the Markdown file that tells the agent what to optimize, what constraints to respect, and what strategies to try.
You set the objective metric. You define the rules. Then you launch the agent and go to sleep. When you wake up, you read results.tsv to see what your agent tried and what worked. You've shifted from experimenter to research architect.