INTERNAL EVALUATION · BY DYMENT.STUDIO
Testing the instrument.
Before external use.
An experimental instrument for checking force and energy consistency in atomic configurations. We are evaluating its numerical reliability, limitations, and ability to produce evidence that can be independently reproduced.
Paid pilots are paused while internal validation is completed.
For a conservative force model, work around a closed configuration path should vanish within numerical error.
Conceptual illustration · not measured data01 / INSPECT THE INSTRUMENT
A finding needs
a reproducible counterexample.
The examples below are saved synthetic reference tests from the experimental 0.1 release. They are not findings about a commercial AI model, and their automated verdicts require independent numerical confirmation.
Recorded findings
A clean result means no inconsistency was detected in these sampled tests. Experimental automated verdicts need independent confirmation; they do not establish physical accuracy, reliability, or universal validity.
Evidence available
Inspect the report data
Your selected report is read inside this browser. It is not uploaded. This viewer does not execute your model.
The internal evaluation compares ordinary point checks, nearby energy-gradient checks, and closed-path evidence. Earlier synthetic benchmarks remain with the experimental runner; they do not establish external-model reliability or an advantage over existing methods.
02 / WHAT AN AUDIT SHOULD EXPLAIN
Evidence to inspect.
Limits to understand.
These are the intended report components under evaluation. Their usefulness for a real model or simulation decision has not been established by the synthetic examples.
- 01
A scoped finding
What failed, where it failed, and whether refinement supports the conclusion.
- 02
The exact counterexample
Coordinates, force outputs, model settings, tolerances, and source identifiers.
- 03
A repeatable regression
A local runner and report your team can use after a model or integration change.
03 / INTERNAL VALIDATION
Internal work continues.
Paid pilots are paused.
We are testing the instrument before accepting paid external work. There is no active pilot intake or payment collection on this page.
Existing verification tools, including OpenKIM, already test energy–force consistency. We have not established novelty, superiority, or readiness for paid external work.
On hold
- Check clean references and deliberate integration errors
- Distinguish numerical artifacts from supported findings
- Replay counterexamples independently
- Compare against established checks at matched cost
- Record unsupported cases and unresolved failures
These are evaluation requirements, not completed validation claims. Paid work remains paused. No model certification, scientific novelty, or simulation-performance guarantee is offered.
04 / EXPERIMENTAL LOCAL RUNNER
Inspect the prototype.
Confirm its results.
The experimental 0.1 Python runner evaluates trusted local model code. Its automated findings require independent confirmation. The website only reads saved reports; it does not run or upload models.
Download experimental 0.1 runner ↓Read the quick start ↗Current boundaries
Experimental scope: position-dependent conservative force components, nonperiodic configurations, and trusted local adapters.
Thermostat forces, friction, external driving, periodic-cell workflows, remote APIs, and unsupported model environments are outside this evaluation release.
Synthetic references provide limited implementation evidence. Independent validation is in progress; accuracy, performance on your model, and readiness for paid work are not established.