How the score works
A transparent way to compare documented AI use cases. The source remains the evidence.
Brain Score
Five values, each from 0 to 2: impact, evidence, reproducibility, maturity and reach. Add them together, then apply a minimum of 1. The highest possible score is 10.
Math.max(1, impact + evidence + reproducibility + maturity + reach)A score is displayed as "7/10 brain cells". It is an editorial reading aid, not a product benchmark or a guarantee of results.
The five criteria
Impact
- 0/2
- No concrete benefit stated.
- 1/2
- A concrete qualitative benefit is described.
- 2/2
- A quantified benefit or material change is reported.
Evidence
- 0/2
- An unsupported claim or thin source.
- 1/2
- A named source with a concrete account.
- 2/2
- Logs, artifacts, measurements or a detailed case support the account.
Reproducibility
- 0/2
- Too little detail to repeat.
- 1/2
- The workflow and tools are described.
- 2/2
- Steps, code or configuration make replication practical.
Maturity
- 0/2
- An idea or exploration.
- 1/2
- A prototype or pilot.
- 2/2
- An application used in real work.
Reach
- 0/2
- A narrowly specific context.
- 1/2
- Useful across similar roles or teams.
- 2/2
- Applicable across several roles, industries or contexts.
From source to entry
Facts are extracted by a model from public web sources. Summaries are AI-assisted using OpenAI Codex. Every entry links to its source and records the source title, type and publication date when available.
The five component values come from this assessment. The formula is implemented in code and recalculates the total at build time. Values are constrained to the integers 0, 1 and 2. The imported brain_score field does not override the formula.
Draft entries have not yet been human-reviewed. The badge stays visible in the catalogue and on each detail page. Human review can change the facts, summary or scores.
Limits and dates
A reported saving is attributed to its source. It is not treated as an independently verified result. A high score does not establish causation, rule out bias or prove that a workflow fits another organisation.
The data date records the inclusion of the supplied dataset in this collection. Source publication dates refer to the original pages. Repository fetch dates refer to API observations. A new website build alone does not change the data date.
Repository rankings
All-time ranks the top 200 repositories in the collected topic sample by current recorded stars. The topics are ai, llm, agents, mcp, claude-code, copilot, rag and generative-ai. Each request returns up to 100 repositories. Results are merged by stable GitHub ID.
Trending compares the latest daily snapshot with the closest available snapshot at least 7 days earlier. Absolute star change is shown only for repositories present in both. Missing history is never counted as zero stars. The displayed dates state the actual comparison window.
Before enough history exists, Trending shows the newest fast risers: the newest 50 repositories in the retained sample, ranked by lifetime stars divided by days since creation, with a minimum denominator of one day. This proxy is not a recent growth measurement.
Failed requests preserve existing data. A dated snapshot is written only after all eight topics succeed. This hobby site does not claim complete GitHub coverage.
Corrections
Send corrections with the entry URL and the supporting source to hello@aiusecaseshub.com.
Browse the collection