THE VERIFIED AI RESEARCH OS
VastiLab builds open-source, audit-native AI research tools for researchers, research groups, and institutions. We help accelerate the path from literature review to research ideas, draft papers, and pre-submission review while keeping sources, decisions, and changes traceable.
AI can accelerate research, but conventional AI-assisted workflows often make provenance, validation, and human accountability difficult to verify. Our projects make evidence, deterministic checks, and explicit handoffs to researchers part of the workflow rather than an afterthought.
Turns collections of papers and slides into a structured, source-grounded wiki, a reverse-linked knowledge graph, and evidence-checked research idea cards. It supports research and course workflows in Claude Code and Codex, including local or remote GPU OCR.
Stage: Active open-source release
View paper-wiki on GitHubTransforms a research idea, proposal, or dataset into a compiled draft paper through a composable workflow. The system checks citations, produces editable vector figures, runs feasible experiments, and applies deterministic integrity gates to keep claims and reported results traceable.
Stage: Active open-source release
View spark-to-paper on GitHubStress-tests a paper before submission through a review, adjudication, revision, and verification loop. Safe editorial fixes are separated from issues that require new evidence, experiments, private knowledge, or the author's scientific judgment.
Stage: Stable open-source releases for Claude Code and Codex
The core tools are publicly available as actively maintained, MIT-licensed open-source releases. Together they form the foundation of The Verified AI Research OS. A hosted product for collaborative research workflows is planned.
Explore all repositories