overview

Teammately Correctness Infrastructure Manual

Learn the correctness infrastructure for scaling expert judgment across coverage, cases, evaluation, and improvement.

Correctness infrastructure

Make specialist AI trustworthy with expert judgment, scaled by AI agents.

Teammately connects Coverage Engineering, Correctness Elicitation, Weave, Trialground, and Coevolve into one inspectable development system.

Start learning Run the first loop

correctness loop

Coverage Engineering
  -> Correctness Elicitation
  -> Weave
  -> Trialground
  -> Coevolve
          -> stronger candidate behavior
          -> sharper coverage and correctness

New reader Learn the product Start with the category, lifecycle, and product boundaries before diving into workflows. Operator Complete a task Follow exact procedures for project setup, contributions, coverage, evaluations, and improvement. Developer Look up objects Find source-labeled object behavior, states, permissions, metadata, and glossary entries. Agent Read machine context Use bounded llms, manifest, search, context, task map, object map, and recovery surfaces.

Artifact map

Follow the correctness loop

Open product loop

01 Coverage engineering Dimensions, Project Topics, construction patterns, representation 02 Correctness elicitation Expert contributions, policies, applicability, rubrics 03 Weave Cases, variants, materials, challenge sets 04 Trialground Harnesses, Run Groups, repeated sampling, Arena, rubric evidence 05 Coevolve Goal Contracts, candidates, evaluations, frontiers

Browse

Jump by workflow stage

All docs layers

Make specialist AI trustworthy with expert judgment, scaled by AI agents.

Follow the correctness loop

Jump by workflow stage

Start here

Operate

Recover