ejentum-reasoning

Causal analysis validates conclusions by routing tasks through specific reasoning dimensions like simulation, abstraction, and metacognition. This structured injection shapes analytical thinking with constraints and blockers to prevent category errors and reasoning drift.

Planning with dependencies requires structured reasoning that applies temporal, spatial, and simulation checks to model multi-step consequences. This approach blocks common failure modes and verifies consistency to prevent missing hidden risks in complex analytical work.

Metacognitive checks are needed for tradeoff evaluation when first-pass answers are likely to miss the true issue. They enforce abstraction and consistency verification to prevent reasoning drift and incorrect conclusions in complex analytical work.

Automated reasoning should be limited to tasks where the first answer is likely to be wrong. It requires a POST request to a specific endpoint with an authorization bearer key, task query, and reasoning mode, meaning it cannot function offline or without API access.