Ussyverse Brainstorm

Research notebook for discovering practical, source-checked ways to combine projects from the Ussyverse organization and mojomast into useful new products.

The Ussyverse contains hundreds of focused experiments. Many have a useful typed model, deterministic transformation, or report hidden behind a prototype interface. This repository asks a simple question:

Which projects become substantially more useful when they share one real workflow, data model, and product surface?

This is a research and planning repository. It does not currently contain the source code of the projects it discusses.

Current recommendation

The strongest next build is Agent Fix Lab, a local tool that turns a failed coding-agent run and a human correction into a reproducible regression case. Start with:

• ussyverse/triageussy for extracting structured error evidence.
• ussyverse/petrichorussy for configuration snapshots and diffs.
• mojomast/hermes-correction-aware-learning for correction records, retractions, and recurrence checks.

The first version should target Python projects and one agent workflow. A Hermes import adapter is follow-on work. AgentReplay is deferred until its TypeScript store is repaired and its architecture is tested.

See the independent product assessment for the decision, runtime evidence, alternatives, acceptance criteria, and proposed two-week validation cycle. The previous shot-to-sound recommendation remains the strongest creative experiment.

Strong candidates

Rank	Product idea	Projects	Primary user
1	Agent Fix Lab	triageussy + petrichorussy + mojomast/hermes-correction-aware-learning	Coding-agent maintainers
2	Service-improvement desk	andoniaussy + shewhartaussy + a3viaussy	Small repair and IT service teams
3	Shot-to-sound capture planner	kuleshovussy + foleyaussy; chromascriptussy later	Independent creators and small film teams
4	Scope-and-capacity quote worksheet	bidussy + taktussy; recapturaussy only if needed	Small service businesses

Documents

• docs/research/conclusion.md: latest independent product assessment, source and runtime findings, ranked recommendations, and next-cycle validation plan.
• docs/COMPOUNDED_IDEAS.md: canonical synthesis of 20 independent brainstorming passes. It reduces 188 raw proposals to 146 distinct candidates and ranks the strongest 30.
• docs/SOURCE_CHECKED_COMBINATIONS.md: prior independent shortlist. All 459 catalog entries were screened and 27 promising repositories received targeted static source inspection.
• docs/REPOSITORY_INDEX.md: complete working index of 330 public ussyverse repositories and 129 public mojomast repositories, captured on 2026-09-13.
• docs/INITIAL_COMBINATIONS.md: the first broad ecosystem clustering and twelve speculative combinations. Preserve it as brainstorming history; later source inspection rejected or narrowed several ideas.
• AGENTS.md: instructions for agents continuing this research.
• docs/ideas-01-developer.md through docs/ideas-20-red-team.md: raw, lens-specific brainstorming passes retained for evidence and ideas outside the top 30.

What makes a good combination?

A strong proposal should have:

1. A specific user with a recurring, recognizable problem.
2. A concrete input, transformation, and useful output.
3. Complementary projects rather than duplicate engines or stacked dashboards.
4. A clear shared data boundary between components.
5. An MVP that can prove value without building a platform first.
6. Source evidence for every reused capability.
7. Honest notes about missing adapters, demo data, unimplemented claims, and validation limits.

Combining projects is not the goal by itself. The combination must make a real task easier than either project would independently.

Research status

The catalog was produced from shallow clones and README/manifest inspection by parallel agents. A second independent pass screened every catalog entry and then inspected source in 27 repositories. The latest assessment inspected 13 candidate repositories, ran four Python suites with 404 passing tests, and used targeted runtime probes to test several important component boundaries. It did not build an integrated product or conduct customer research.

Repository descriptions are leads, not proof. Before implementing an idea, pin revisions, run each component's tests, exercise it with user-authored data, and verify that its actual interfaces support the proposed handoff.

Contributing

Open an issue or pull request with a proposal that includes:

• Exact repository names and source links.
• Target user and recurring problem.
• Input to processing to output flow.
• Existing code that can be reused.
• Adapters and product work that do not yet exist.
• A small demo and acceptance criteria.
• Risks, maturity caveats, and rejected alternatives.

Agents should read and follow AGENTS.md before adding or changing recommendations.