Roadmap

Roadmap

0.1: markers in notes, configs, scripts and the paper; results files; the map, the diff and warnings; checkpoints, the live page, the Claude Code hook, setup claude and the MCP server.

Next

• MLflow (its local mlruns store) and W&B runs as results, tagged with an experiment id.
• research-diff suggest: your LLM proposes markers for untagged notes and papers, shown as a change you accept; it never writes them itself.
• Claims checked against numbers: expects: acc > 0.85 on a claim, and a warning when a result no longer meets it.
• A Claude Code plugin, and PyPI releases.