Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Research pipelines as semantic execution units: each skill declares inputs/outputs, acceptance criteria, and guardrails. Evidence-first methodology prevents hollow writing through structured intermediate artifacts.
| Date | Stars |
|---|---|
| 2026-07-24 | 490 |
| 2026-07-25 | 492 |
| 2026-07-28 | 493 |
| 2026-07-30 | 493 |
| 2026-08-06 | 493 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# Research Harness Turn a research goal into a reviewable deliverable while preserving the sources, decisions, intermediate artifacts, and execution evidence behind it. Research Harness is an end-to-end **Auto Research Design System** built from two complementary parts: - **Skills** perform bounded research transformations such as retrieval, extraction, comparison, synthesis, review, and writing. - **Harness** organizes those Skills into recoverable Workflows, checks their artifacts, records what happened, and locates the next repair when a Run fails. ```text Goal -> Run -> Evidence -> Improve ``` The project does not claim to be an autonomous scientist. It makes long-form research work observable and correctable so a user or agent can deliver, audit, resume, and improve it without reconstructing the whole process from chat history. ## Choose The Outcome Users choose a Workflow by the result they need. The internal Skills and Units remain implementation details until inspection or repair is necessary. | Desired result | Workflow | Starting point | Main deliverable | |---|---|---|---| | Understand a topic and decide what to read | `research-brief` | topic | `output/SNAPSHOT.md` | | Review one paper or manuscript | `paper-review` | manuscript | `output/REVIEW.md` | | Synthesize studies under an explicit protocol | `evidence-review` | review question, then an approved protocol | `output/SYNTHESIS.md` | | Write a literature survey or bounded research report | `arxiv-survey` | topic and delivery constraints | `output/DRAFT.md` | | Deliver the same Survey path as LaTeX and PDF | `arxiv-survey-latex` | topic and delivery constraints | `latex/main.pdf` | | Develop literature-grounded research directions | `idea-brainstorm` | topic and scope | `output/REPORT.md` | | Turn an existing source set into a tutorial | `source-tutorial` | source pack and audience | tutorial, article PDF, slides | `graduate-paper` remains a research-stage Chinese thesis path. It contains useful Skills but is not part of the seven executable Pipeline contracts. The input boundaries are deliberate. `research-brief`, the Survey family, and `idea-brainstorm` can begin from a topic. `paper-review` requires a manuscript; `source-tutorial` requires a local source pack and audience; `evidence-review` turns a review question into a protocol and stops for approval before retrieval. The Harness names a missing prerequisite instead of silently replacing it with invented context. ## Start A Run The CLI currently runs from a source checkout, requires Python 3.10+, and uses [uv](https://docs.astral.sh/uv/). `uv run` installs the declared Python dependencies, including PDF manuscript extraction. Source Tutorial PDF ingest also requires `pdftotext`; LaTeX/PDF delivery requires the TeX tools named by the selected Workflow's compile checks plus either Poppler's `pdfinfo` or the optional `PyMuPDF` package for page-count acceptance. Goal-seeded Workflows can then start: ```bash uv run rh goal create \ --goal "Understand test-time adaptation for robotics and decide what to read" \ --workflow research-brief \ --workspace workspaces/robot-adaptation uv run rh run start --workspace workspaces/robot-adaptation uv run rh run status --workspace workspaces/robot-adaptation uv run rh run approve --workspace workspaces/robot-adaptation --checkpoint C2 uv run rh run resume --workspace workspaces/robot-adaptation uv run rh evidence inspect --workspace workspaces/robot-adaptation --excerpt ``` `run start` advances until the next unmet prerequisite. For `research-brief`, inspect the core paper set, taxonomy, outline, and C2 review block before approving C2; only the currently active Checkpoint can be approved. The approval is bound to the reviewed Artifact hashes, so changing those files invalidates stale authorization. `run resume` then continues from the persisted Unit ledger. The completed Run contains a readable brief and structured scorecard. `evidence inspect` wr
Excerpt of 12,769 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:ee5ee7f96fc19cfa, topic:gpt
matched fp:ee5ee7f96fc19cfa, topic:vibe-coding