KeilerHirsch-Labs is the home for projects that have grown beyond personal experiments and deserve their own engineering, documentation and release discipline.

The common thread is simple: measure first, preserve evidence, minimize hidden trust, and automate only what is understood well enough to automate responsibly.

Research-first framework for reproducible, uncertainty-aware and auditable measurement of AI-system behaviour.

BRONCO starts with metrology and experimental validity rather than a leaderboard: measurands, construct validity, repeatability/reproducibility, uncertainty, provenance and evidence come before benchmark features.

Its foundation is informed by DIN/ISO/IEC and JCGM/BIPM metrology work, including DIN EN ISO/IEC 17025, ISO 5725, ISO/IEC 25059, ISO/IEC 5259 and related AI standards. A deliberately small Ada/SPARK trusted core is used for measurement-critical deterministic logic and formal verification.

Status: Foundation / Research · License: MIT

Standards references describe engineering alignment and traceability. They do not imply DIN/ISO/IEC affiliation, certification or laboratory accreditation.

Local-first export and synchronization tooling for your own Claude data.

A single Go binary that exports your own claude.ai conversations, project documents and memory to local Markdown. The project is Windows-focused, uses DPAPI + CDP, and is designed without telemetry or a cloud dependency for the exported archive.

Status: Early Access · License: AGPL-3.0

The roadmap includes deeper local knowledge/persistence workflows, including integration work around MemPalace.

• Evidence before claims — raw observations, provenance and reproducible procedures matter more than screenshots or anecdotes.
• Research before features — measurement systems need valid constructs and procedures before dashboards and rankings.
• Local control where practical — user-owned data should remain inspectable and portable.
• Small trusted computing bases — high-assurance logic is isolated instead of pretending an entire rapidly changing stack can be trusted equally.
• Tests and review before release — CI, explicit licensing and adversarial review are expected project hygiene.
• Formal methods where they buy real assurance — contracts and proofs are used for critical deterministic logic, not as decoration.
• No silent uncertainty — limitations, assumptions and unresolved ambiguity belong in the engineering record.

Issues are the most reliable public entry point for bug reports, research criticism, reproducibility findings and technical proposals. For BRONCO in particular, a strong counterexample or standards correction is currently more valuable than a feature request.

_{Org avatar built from five icons by Lorc via game-icons.net ("Stag Head", "Boar Tusks", "Erlenmeyer", "Test Tubes", "Round Bottom Flask"), recolored and recomposed, CC BY 3.0. Full credits in CREDITS.md.}