Generate your submission with ChatGPT

Every PWM submission is two files: a human-readable Markdown and a machine-readable JSON (Solutions add code/ + reports/). You only write the Markdown — your LLM generates the rest from the rules below.

Download a template

Grab the MD + JSON for your layer, or fork an existing artifact close to yours.

Polish in your LLM

Paste the template + the prompt below into ChatGPT / Claude / Gemini. Edit the prose; let the LLM regenerate the JSON.

Submit the bundle

Upload your MD + JSON at /submit. Founder review promotes it to mainnet — your wallet earns PWM.

Download templates

L1 Principle ⬇ principle.md ⬇ principle.json

L2 Spec ⬇ spec.md ⬇ spec.json

L3 Benchmark ⬇ benchmark.md ⬇ benchmark.json

L4 Solution ⬇ solution.md ⬇ solution.json

Or open any artifact in the catalog and click Download bundle to fork it.

📋 The prompt — copy into your LLM

You are helping me submit an artifact to the PWM Protocol (Physics World Model).

I will paste a Markdown template. Your job:
1. Help me rewrite the Markdown so the science is correct and clearly explained.
2. Then regenerate the sibling JSON file so EVERY field matches the Markdown.
3. Preserve the schema documented in the "File Mapping" footer at the bottom of the Markdown.
4. Output the Markdown and the JSON each in its own fenced code block, tagged with the filename.

Rules:
- The Markdown is the source of truth. The JSON must never contradict it.
- Keep all section headings from the template; fill them in for my artifact.
- Use SI units. State the forward model as an equation. List operator primitives.
- Do NOT invent benchmark scores. Leave performance numbers blank unless I provide them.
- For an L4 Solution: also produce code/solver.py and reports/reproduce.md, and remind me that reports/results.md must come from ACTUALLY RUNNING the solver — never fabricated.

Here is my template:
[PASTE THE .md FILE HERE]

What each bundle must contain

Layer	Files	Must include
L1 Principle	.md + .json	Forward model, physics fingerprint, operating range Ω, tolerance ε
L2 Spec	.md + .json	Parent principle ID, domain Ω with bounds, tolerance ε function
L3 Benchmark	.md + .json	Parent spec ID, dataset URL/IPFS link, metric + threshold, baseline scores
L4 Solution	.md + .json + code/ + reports/	Parent benchmark ID, solver code, and a reproduction report from actually running it

The mapping between the MD and the JSON (and code) is written at the bottom of every template MD, in the File Mapping section — that is what your LLM uses to keep them in sync.

Bundle ready? Submit it and start earning PWM.

🚀 Go to /submit →