How hbar.science works
What is an artifact?
An artifact on hbar.science is a structured pointer to a scientific object, combined with a manifest that records its epistemic status. hbar.science does not host the artifact itself — it hosts the metadata: what the artifact is, what it claims, how reproducible it is, and what AI involvement it had. The artifact lives elsewhere (a repository, a preprint server, a dataset archive); the manifest makes it inspectable and traceable here.
Large executables, raw data files, and continuously evolving assets are not hosted on this platform. Only links, metadata, and structured records are maintained here.
Examples of artifacts
- —A Python implementation of an optimization algorithm, hosted on GitHub, pointed to from a code manifest
- —A parameter sweep analysis of regularization effects in VQE, with results in a notebook, pointed to from an analysis manifest
- —A synthesis essay on structural constraints in scientific publishing, hosted as a page on this site, pointed to from an essay manifest
The labeling system
Every artifact is assigned three independent labels. They describe different dimensions of the artifact and do not imply each other.
Zone — type of scientific act
What kind of epistemic act produced this artifact?
- Hypothesis — a testable conjecture or theoretical prediction, without experimental backing
- Exploration — preliminary investigation, proof-of-concept, parameter scan
- Evidence — empirical result with reproducibility guarantees and error analysis
- Synthesis — integration of findings, review, conceptual framework
Tier — validation depth
How reproducible and validated is this artifact?
- T0 — concept sketch; at least one claim is documented
- T1 — runs and produces output; deterministic seed documented
- T2 — internally replicated; replication status confirmed within the same team
- T3 — independently replicated; environment hash documented
AI Level — degree of AI involvement
How much did AI systems contribute to producing this artifact?
- A0 — no AI used
- A1 — AI used for mechanical tasks (editing, formatting)
- A2 — AI used for cognitive assistance (code generation, analysis, hypothesis refinement)
- A3 — AI generated core artifacts under human supervision
- A4 — autonomous AI; full scientific loop run by AI
These three axes are independent. A T0 artifact can be Zone: Evidence. An A3 artifact can be Tier: T2. Labels describe structure, not quality.
Submission workflow
- 1.
Generate a manifest at /submit
Fill in the form. The tool computes a recommended Zone, Tier, and AI Level based on your inputs, flags any compliance gaps, and generates a manifest.json file.
- 2.
Submit via GitHub issue or pull request
Use the prefilled GitHub issue link generated by the submit form, or open a pull request directly adding artifacts/{slug}/manifest.json to the repository.
- 3.
Review → merge → live
Submissions are reviewed for schema validity and classification consistency. Once merged, the artifact appears on hbar.science automatically.
Review model
What is evaluated
- — Schema validity (all required fields present)
- — Classification consistency (Zone/Tier/AI match the declared reproducibility data)
- — Tier compliance (declared tier requirements are met or mismatches flagged)
- — Slug uniqueness and format
What is not evaluated
- — Scientific correctness of claims
- — Validity of methods or experimental design
- — Significance or novelty of results
- — Quality of the external artifact being pointed to
Limits
- —No medical, clinical, or safety claims are hosted on hbar.science.
- —Large executables, datasets, or continuously evolving assets are not hosted here — only links and structured metadata.
- —Inclusion on hbar.science does not constitute peer review, endorsement, or verification of scientific correctness.
- —Contributions are currently by invitation or via the submission workflow. Access to publishing is not open by default.