Kenton Labs / Research / Formal systems
Formal
Computer
Science.
Precise problems. Inspectable evidence. Mechanically checked answers.
Shortest path / KL-FCS-028 / Distance 5A path proves feasibility. Complete comparison establishes the minimum.
A structured knowledge base for computation, languages, programs, optimization, and mathematical systems.
Explore the areas.
Each dataset connects definitions, a formal model, worked instances, acceptance rules, complexity, failure cases, and references. Open an area for its complete exposition and evidence.
Work through the mechanics.
Change inputs and follow the computation. Signals, states, paths, schedules, and exact fractions make the rules visible. All interactions run locally in your browser.
Inspect the records.
Read the exact problem and candidate result, then inspect the evidence and checker reasoning. Negative results and counterexamples belong in the knowledge base alongside successful solutions.
{{RECORD_COUNT}} of {{RECORD_COUNT}} records
No matching records. Try another term or reset the filters.
An inspectable standard.
A fixed specification defines the truth conditions. The result, its evidence, and the procedure that checks it remain attached to that specification.
Publish the whole argument.
Definitions lead to the model; the model leads to the instance; the claim leads to its evidence. Every record has an explanation, explicit assumptions, input encoding, acceptance procedure, scope, references, and provenance.
Algorithms, automata, programs, and optimization problems may admit multiple valid witnesses. The checker accepts a property or objective, rather than relying on an answer’s wording.
Read the publishing guidelines ↘Name what was checked.
Finite enumeration, exact witness replay, type reconstruction, and rational arithmetic support different claims. Their bounds and trust assumptions remain visible. Feasibility and optimality are assessed separately.
Verification metadata records the checker, source hashes, runtime, and human-review state. Unknown results and timeouts remain unresolved. Corrections preserve the previous source release.
Inspect checker rejection tests ↘Read. Run. Reproduce.
The complete bundle includes the corpus, schema, explanatory entries, area definitions, source references, browser graphics, Python checkers, rejection tests, and release tools.
From the extracted bundle, run python3 tools/verify.py and python3 tools/check_checker.py. The Python interpreter and custom checker implementations form the verification trust boundary. Human review has not yet been recorded. Source is available for inspection and reproduction; reuse-license selection remains pending.
The editorial structure is informed by Algebrica’s publishing approach. All reference instances and illustrations here are original Kenton Labs work; linked sources provide conceptual background. No external dataset or model-performance evaluation is claimed.