Type Someone Now
live loopEvidence triage, candidate board, evidence ledger, adjacent-type duel, next questions, report draft, share link, and JSON recovery.
Start with a live typing session, audit an answer, replay benchmark traps, collect follow-up evidence, or install the same protocol into a mainstream or unknown agent host.
Evidence triage, candidate board, evidence ledger, adjacent-type duel, next questions, report draft, share link, and JSON recovery.
Paste a response or report, check missing runner-up/falsifier/evidence movement, replay hidden references, and generate repair prompts.
Browse benchmark traps, top-two conflicts, falsifier themes, blind review packets, and public-safe follow-up observations.
Generate pack commands, install checklists, adapter receipts, and issue seeds for Codex, Claude Code, Cursor, opencode, Copilot, Windsurf, Cline, Continue, aider, and more.
Describe an unknown host, map its real instruction surfaces, and export a Universal Agent Bridge plan without forking the canonical protocol.
Turn a miss, benchmark trap, answer-quality failure, or consented follow-up observation into a structured GitHub issue.
The same safety contract runs through every route: candidate set, serious runner-up, evidence ledger, falsifier, and safety boundary.
Use this when you want an agent to run the same protocol instead of guessing a label.