Three-Gate Research Protocol

grounded discovery engine

Loading framework…

Three-Gate
Research Protocol

A rigorous framework for objective, grounded discovery.

What is this?

The Three-Gate Research Protocol is a structured methodology for evaluating scientific and investigative claims. It uses a strict temporal sequence — Generate, Gate, Verify, Find the dropped thread, Grade, and Write — to ensure that creative exploration does not compromise empirical rigor.

Central to the system is a triple-filter validation process that checks claims for novelty (Gate 1), independent corroboration (Gate 2), and underlying mechanism (Gate 3) — separating high-confidence “bets” from mere folklore.

How it works

1
Generate WideExplore broadly without filtering — the goal is surface area
2
Three GatesSilence/Novelty → Reproducibility/Corroboration → Mechanism/Provenance
3
Sideways LookAsk what the standard account wouldn’t predict
4
Find the Dropped ThreadHunt abandoned connections — they carry the highest value
5
Grade ClaimsTag each finding [C]onsensus, [D]isputed, or [H]ypothesis
6
Attackable OutputWrite the version that survives checking — not the triumphant one

How to use it

Type a research question or paste a document below. The protocol runs the full pipeline against the model you select — you’ll see live reasoning, gate scores, evidence levels, and graded claims appear in real time.

? Try “Is there evidence for cold fusion?”
? Attach a PDF for document-grounded analysis
? Switch modes: Discover (science) or Investigate (people & claims)
Verify ON

Three-Gate Research Protocol

Run the protocol to see backend processing in real time…
idle

Silence / Novelty

—
Score: — Pending
Gate 1 awaits protocol run.

Reproducibility / Corroboration

—
Score: — Pending
Gate 2 awaits protocol run.

Mechanism / Provenance

—
Score: — Pending
Gate 3 awaits protocol run.

Evidence timeline

Gate-level evidence breakdown
Gate 1 · Silence / Novelty
—
Gate 2 · Reproducibility
—
Gate 3 · Mechanism
—
Supporting evidence will appear here when the protocol runs.

Findings panel

Graded claims
[C] Consensus · [D] Contested · [H] Hypothesis

Documents & settings

Recent runs