Get Physics Done launches open-source physicist
Physical Superintelligence PBC’s Get Physics Done is an open-source AI copilot for physics research that turns a problem into a structured workflow: scope, plan, derive, verify, and package. It’s aimed at long-horizon, math-heavy work where brute-force prompting falls apart.
This feels like one of the more credible takes on an “AI scientist” because physics forces the agent loop to prove its work, not just sound smart.
- –It targets real research workflows, not chatty one-off answers, which makes the product much more substantive than generic agent demos.
- –The built-in emphasis on verification is the right instinct for physics, where a plausible derivation is often not enough.
- –Runtime support across Claude Code, Codex, Gemini CLI, and OpenCode should lower adoption friction for researchers already living in terminals.
- –Open source matters here: people can inspect the workflow, modify it, and judge whether the agent is actually useful in hard cases.
- –If it holds up, this could be a useful template for other scientific domains that need structured reasoning and reproducibility.
DISCOVERED
69d ago
2026-03-21
PUBLISHED
69d ago
2026-03-21
RELEVANCE
AUTHOR
Github Awesome
