AI AUTOPILOT · EVIDENCE GATES · HUMAN TAKEOVER

Make AI-assisted social-science research auditable.

Codex Research Stack v0.6 Experience Preview starts from a real local project, then turns research setup into a Cockpit with visible setup progress, Autopilot runs, AI suggestion cards, evidence gates, method artifacts, recovery actions, project memory, diagnostics, and auditable reports.

Materials to claims to evidence to quality checks to status file

Getting started · Research Autopilot · Installation · Integrations

Real launch flowCreate a local project, confirm a method route, and seed the first material and claim before entering Cockpit.

Research CockpitSee setup progress, current state, gate board, memory, recovery queue, timeline, AI suggestion source, and export readiness.

Suggestion cardsAccept, edit, or reject AI suggestions before project facts change.

Evidence gatesBlock unsupported claims and incomplete evidence before writing.

Method workflowsCompose routes, modules, and user-confirmed method artifacts.

Run reportsExport what AI did, what the user accepted, and what remains blocked.

DiagnosticsRun project health checks and export redacted diagnostic snapshots for support and release review.

Local-firstUse the project workflow without Codex, Python, Zotero, Obsidian, or an API key.