ThinkTools Protocol

The Primitives

Most AI failures stem from asking a model to do everything in one pass (analyze, plan, execute, critique). ThinkTools decouple these cognitive operations into distinct, specialized modules.

1. GigaThink (Divergent)

Goal: Generate maximum variance in perspective before convergence.

Method: Forces the model to adopt 10+ distinct personas (e.g., "The Skeptic," "The Optimist," "The Historian") to analyze a problem.

use reasonkit::thinktools::GigaThink;

let thinker = GigaThink::new().depth(10);
let perspectives = thinker.expand("Should we rewrite in Rust?").await?;

2. LaserLogic (Deductive)

Goal: Validate the logical consistency of a proposed plan.

Method: Decomposes arguments into syllogisms and checks for formal fallacies.

use reasonkit::thinktools::LaserLogic;

let logic = LaserLogic::new();
let flaws = logic.audit(&plan).await?;

3. ProofGuard (Verification)

Goal: Prevent hallucinations via multi-source triangulation.

Method: Requires 3 independent citations for every factual claim. (Currently in simulation mode for Core).

use reasonkit::thinktools::ProofGuard;

let guard = ProofGuard::new().min_sources(3);
let verified = guard.verify(&claim).await?;

Citation

If you are using ReasonKit in your research or production system, please cite the protocol:

@software{reasonkit_2025,
  author = {ReasonKit Team},
  title = {ReasonKit: Auditable Reasoning Protocols for Production AI},
  url = {https://reasonkit.sh},
  version = {0.1.5},
  year = {2025}
}