What happened
The emlint GitHub repository describes version 0.2 as a preliminary release of a static linter for Stim Detector Error Models. The project’s stated purpose is to find structural problems in a DEM before simulation. Its README contrasts the tool with Stim and sinter: Stim simulates, sinter samples, and emlint verifies properties of the DEM. The repository lists an Apache-2.0 license and says version 0.2 behavior is usable, although internal APIs and implementation details may change before version 1.0. The citation block identifies version 0.2.4.
Users can install the package with pip install emlint and run emlint check circuit.dem. The CLI also accepts raw DEM text, while the Python interface accepts a path, raw text, or a DEM generated from a Stim circuit. Project defaults can be stored in pyproject.toml. The repository describes ten checks, including detectability, sensitivity, observable coverage, probability bounds, duplicate mechanisms, correctability, null mechanisms, high-probability mechanisms and decoding-graph connectivity. Boundary symmetry is skipped when fewer than two logical observables are present.
A check report can identify the mechanism, detector or observable involved. In the included distance-5 surface-code example, the buggy DEM has a detector, D17, that no error mechanism triggers, and the decoding graph has a second component containing only detector 17. The linter marks these findings as warnings in that example and gives a hypothesis about a missing noise instruction or a redundant detector. The README says the example passes after D17 is reconnected to its neighbors. It describes findings as taking milliseconds, in contrast to a silent DEM bug that might not surface until one million simulation shots, approximately 45 minutes.
The tool supports text, JSON and SARIF output, selectable checks with --only and --ignore, and differentiated exit codes: 0 for all checks passing, 1 for an error-severity finding, 2 for warnings only, and 3 for unlintable input. The repository gives a GitHub Actions example that installs emlint and runs it over .dem files, as well as a pre-commit hook that checks staged files at error severity. Warning-only results do not block that hook; input failures still return code 3. The project also documents an agent integration guide covering exit-code handling, its JSON contract and context declaration.
What the archive adds
New Runtime’s earlier coverage on 21 October 2026 described Claude Mythos 5’s reported quantum-circuit results as notable for using a closed optimization loop with externally measured correctness and resource use (earlier coverage). emlint is a new tool event in the same broader area of making quantum-computing results checkable, not a report of a new model result.
Trend signal
This is a modest example of the value of inspectable evidence in scientific computing: the tool reports specific structural properties and counterexamples rather than presenting a single opaque pass/fail claim. That is related to the emphasis on checkable artifacts in New Runtime’s scientific-agent pattern. The connection is limited: emlint is a DEM linter, not an autonomous scientific agent or a complete evidence package.
Why it matters
For teams building quantum error-correction workflows, a structural error in a DEM can waste simulation time or distort the interpretation of logical error rates. A fast local check can catch some classes of mistakes earlier, and machine-readable results make it practical to run those checks in CI or an agent-assisted fix loop. The counterexample matters as much as the check: naming a detector or mechanism gives a researcher somewhere concrete to investigate instead of leaving them with a generic warning.
The boundary is equally important. emlint checks a DEM, not whether that DEM faithfully represents the source circuit. A team can therefore use it as one gate in a layered validation process, but should retain circuit-level checks with Stim and use a decoder or simulation for the questions those tools answer. Treating the linter as a proof of physical or circuit correctness would exceed its stated scope.
Limits and uncertainty
The repository calls v0.2 preliminary and warns that internal APIs and implementation details may change before v1.0. It does not establish independent adoption, performance across a broad set of real-world DEMs, or how often the checks catch consequential defects in practice. Some checks are heuristic warnings, and the README notes that decomposed DEMs may trigger a duplicate-signature warning that is expected rather than a concatenation bug. Most importantly, the tool’s stated guarantees concern DEM structure only: they do not establish circuit semantic correctness or decoder success.
Skill to build now
If your team maintains .dem files, try emlint on a small fixture and inspect both its warnings and exit codes before adding it to CI. Start with error-severity checks on changed files, route JSON or SARIF into existing tooling, and make sure any automation treats exit code 3 as an input failure rather than as findings about the model.
Reader takeaway
emlint makes several structural checks on Stim detector error models cheap enough to run in a development loop, and it points to concrete counterexamples when checks fail. Use it as a DEM-level linting gate—not as evidence that the circuit itself is correct.