SIMAT starts from AI in Science and grows toward AI and Science.
AI is a guest at the door: it edits a snippet, finds a paper, and serves science passively.
AI is a resident participant: it runs analysis, coding, compilation, execution and physical validation end to end.
AI and research co-evolve: experience, cases and criteria keep accumulating, so each task makes the next one easier.
The input is one sentence in natural language; the output is a clean-compiled, physically validated Gadget4 variant.
Models usually come with a complete mathematical description: equations, parameters, algorithms, earlier implementations.
What is missing is not the idea, but the people who turn it into correct code.
Getting it right inside a 100,000+ line optimized C++ codebase — across the gravity solver, the time integrator, particle updates and I/O.
The same job under four working modes, measured in real hours; bar length is linear in manual effort.
Every stage leaves an artifact; every failure has a way back.
It does not just raise an error; it works backwards from the deviation like an investigator.
Hand-written specs × automatic accumulation: seed skills are injected by experts, grown skills emerge from real tasks.
Each case ran the full pipeline from one natural-language prompt; every deliverable passed physical validation.
Transfer-function cutoff plus thermal velocities, settled by IC-level validation.
Monte Carlo scattering; the SIDM signature is reproduced.
A decades-old theory, implemented in an N-body code for the first time.
Same binary, same parameters, four consecutive 256³ IC generations.
12,779 reports contain only 2,156 scattering pairs in total; physics expects 10⁵–10⁶.
The 32³ first pass reported “0 scattering pairs” as if it were the expectation; the P0 defect only surfaced after a 256³ run of 45 hours. It is now frozen as criteria #033 / #037.
The anisotropy of Bianchi I is not a slogan: it leaves the same directional imprint on the expansion history, the power spectrum and the velocity distribution.
Analysis belongs to the Python multi-agent layer, actions to Skills; one /simat-master orchestrates both.
Analytical work: literature search, code analysis, spec and report generation.
Direct actions: editing, compiling, running simulations, distilling experience; the instructions are themselves executable specs.
It is powerful, but not omnipotent; clear boundaries show where to go next.
It cannot invent an algorithm the literature has not yet pinned down; it can only implement what has already been described clearly.
With incomplete criteria, the gate may accept wrong results or reject correct ones. Maintenance by domain experts remains essential.
The architecture is code-agnostic; moving SIMAT to Arepo, Enzo or RAMSES only requires retraining the code analysis.
MIT / GPLv3 · questions, reuse, modifications and derivative work are all welcome.
docs · demo · 10 cases · 16 skills · GADGET-4 reference source
This is a 16:9 deck; it scales to fit any screen and prints landscape at 13.33 × 7.5 inches.
Printing to PDF from Edge is recommended; the print styles are built in, so no extra layout work is needed.