For you
Developers
Name judgments, bound units, and wire deterministic code around cognition.
Cover for Developers12 concepts curated for developers.
Back to home
Illustration for Cognitive UnitsThe bounded home for open judgment
9 minCognitive Units
A cognitive unit is a bounded, callable contract that owns one open judgment, inputs, outputs, authority limits, and evaluation hooks, not a prompt file with a label.
Illustration for Open, closed, and closingTwo questions that sort work
12 minOpen, closed, and closing
Before you pick a model or write a prompt, two sorting questions tell you whether a decision belongs in code, in cognition, or with a person. Most architecture failures start with getting that sort wrong.
Illustration for Where does a decision live?One home for each judgment
7 minWhere does a decision live?
Every open decision should have one primary home in the architecture, not smeared across prompts, UI, and middleware. This page explains why locality errors make systems untraceable and unmaintainable.
Illustration for AgentsOwns progress toward a goal
9 minAgents
An agent owns a goal and runs the minimum sufficient loop, plan, act, observe, review, delegating bounded judgments to cognitive units and skills.
Illustration for Deterministic CodeThe layer that must stay exact
9 minDeterministic Code
Cognition proposes. deterministic code validates, enforces, and acts. Exact work, allergies, portions, permissions, must not remain open to model interpretation.
Illustration for Start with one bounded judgmentShip one call end to end
11 minStart with one bounded judgment
After framing names the outcome, the first construction move is one cognitive unit with a contract, evaluation suite, and declared gaps, not a platform diagram with nothing measurable.
Illustration for Adopting cognitive units in existing codeInventory before framework
10 minAdopting cognitive units in existing code
Most systems already contain unnamed judgments inside prompts and scripts. Adoption names them, shadow-evaluates them, and cuts over incrementally, not by rewriting everything behind a new framework.
Illustration for How decisions closePaths from open to closed
15 minHow decisions close
Decisions move from open to closed as cases accumulate, answers stabilise, and rules become writable. Closing is a lifecycle, not a moment of prompt writing, and cache hit rate tells you where you are.
Illustration for An agent is mostly ordinary codeLoop, state, and calls
9 minAn agent is mostly ordinary code
An agent is not a model with tools. It is largely deterministic control flow in which judgment is concentrated in named cognitive units and known work in Skills.
Illustration for Skill versus capabilitySemantics versus packaging
9 minSkill versus capability
Skill is a defined responsibility boundary. Capability is informal unless mapped to skill, cognitive unit, or library package.
Illustration for Acting is not agencyPerforming is not pursuing
9 minActing is not agency
A Skill may write to the world. Reach does not create an agent. Agency begins when evidence can change what work should be undertaken next.
Illustration for Skill library versus cognitive unit libraryTwo inventories, two cuts
10 minSkill library versus cognitive unit library
Cognitive unit libraries publish bounded judgments. skill libraries publish responsibility-sized procedures, different versioning, evaluation, and substitution semantics.