psi-oss

psi-oss / get-physics-done

Public

The first open-source agentic AI physicist, by Physical Superintelligence PBC (PSI).

www.psi.inc agentic-ai claude codex gemini meta-prompting
72
9
100% credibility
Found Mar 15, 2026 at 72 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Get Physics Done is an open-source AI copilot that structures physics research workflows from problem formulation and planning through execution, verification, and manuscript preparation.

How It Works

1
🔍 Discover GPD

You hear about Get Physics Done, a helpful AI sidekick built by physicists to tackle tough research problems from idea to paper.

2
⚙️ Add to your AI chat

Run one simple command in your terminal to install GPD into your favorite AI coding tool, like Claude or Gemini, so it's ready whenever you chat.

3
🚀 Start your physics project

Tell GPD your research question, like bounding operator dimensions in the Ising model, and it asks smart questions to clarify your goals and assumptions.

4
📋 Plan the work

GPD breaks your project into clear phases with tasks, dependencies, and checks, creating a roadmap you can follow step by step.

5
Do the research
✅
Everything checks out

Results pass verification, ready for the next phase.

⚠️
Spot an issue

GPD flags a gap or inconsistency, so you refine and retry.

6
📄 Write and review

GPD drafts your paper, runs peer review simulations, and helps respond to feedback for submission.

🎉 Physics done!

You have a complete research package with verified results, clean derivations, and a publication-ready manuscript.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 72 to 72 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is get-physics-done?

Get Physics Done is a Python-based AI copilot that installs into tools like Claude Code, Codex, Gemini CLI, or OpenCode, turning vague physics research questions into structured workflows. It handles formulation, planning, execution, and verification—producing artifacts like project roadmaps, derivation notes, verification scripts, and even manuscript drafts for arXiv submission. As the first open-source agentic AI physicist, it's a pioneering GitHub project automating long-horizon physics tasks from the first GitHub commit.

Why is it gaining traction?

It stands out by enforcing physics-specific rigor: dimensional checks, limiting cases, and convention locks across 18 fields, which generic AI prompting can't reliably deliver. With 61 commands like `new-project`, `plan-phase`, `verify-work`, and `peer-review`, it integrates seamlessly into existing AI terminals via simple npx install, outputting real deliverables without manual orchestration. Early adopters praise its first open-source release tackling reproducibility and peer review—rare in AI research tools.

Who should use this?

Theoretical physicists grinding multi-phase projects like conformal bootstraps or RG flows, where verification trumps speed. Numerical researchers needing parameter sweeps, sensitivity analysis, or experiment comparisons. Academics prepping papers for journals like Nature or PRL, especially those tired of fragmented prompting in Claude or Gemini.

Verdict

Promising first open-source project for physics AI, but at 72 stars and 1.0% credibility score, it's raw—docs are solid but expect iteration on edge cases. Try it if you're in physics research; skip for production unless you're pioneering your first GitHub contribution here.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.