vibe-research

To keep AI agent memory and human context aligned during a literature review, use a Markdown-first research OS with shared control files like STATE.md and TASKS.md to maintain a stable, human-readable memory layer for both humans and agents.

The best way to track AI-assisted research tasks and ensure reproducibility is to enforce gate-driven workflows with versioned artifacts, using Git as a historical memory layer to track changes and milestones without conflating runtime state.

To set up a project folder for AI-assisted experiment planning and data analysis, bootstrap a new MD-first research OS using a provided script, which creates shared control files and optional resources directories like scripts/ and references/ on demand.

Yes, you can use Markdown files to manage the entire research lifecycle from hypothesis development to writing by utilizing artifacts like AGENTS.md and CHANGELOG.md as a shared control layer, ensuring clear handoffs and traceability across all research phases.

Git works effectively as a historical memory layer for coordinating literature triage and evidence synthesis, tracking changes and milestones independently from runtime state to keep human and agent memory aligned within a single project folder.

A limitation of using a Markdown-first research OS for project management is that it relies on developers manually maintaining structured artifacts like AGENTS.md and STATE.md, requiring strict adherence to gate-driven workflows to prevent context drift between human and agent memory.