KCC 00.1

Executive Summary

KCC makes agentic delivery governable by replacing informal prompting with repeatable lifecycle artifacts and measurable quality gates.

IntroductionKernelCapabilitiesCellsPhases
Created 2026-06-08 · v0.4.0

The Problem

Agentic systems generate output faster than human organizations can absorb it. Cost scales with usage. Decisions accumulate without traceability. Learnings stay trapped in the team that discovered them.

AI adoption produces a structural mismatch: generation compounds; organizational learning does not. Without architectural intervention, organizations end up with sophisticated agents producing untraceable, ungoverned, fragmented output at scales their human systems cannot absorb.

The Answer

KCC is an operating model for organizations choosing to compound learning rather than accumulate fragmentation. It rests on three layers.

  • Kernel — a small, slow-changing shared contract every agent honors. Specification, not runtime. Owned by 3-7 named maintainers.
  • Capabilities — versioned, reusable agents in a shared catalog. Mature through L1 (Experimental) → L2 (Proven) → L3 (Golden Path).
  • Cells — team-owned compositions of kernel + capabilities + custom agents. Where work ships.
The kernel does not run the cells. It defines the contract they all honor.

Two meta-agents enforce the contract.

  • Token Guard — declared cost envelopes, enforced per invocation.
  • Butler — gating decisions (HITL vs HOTL) based on demonstrated behavior, not role.

A nine-surface agent contract governs every agent. The last two surfaces — Confidence and Decision Trace — form a cognitive substrate that makes agent decisions auditable, replayable, and patternable.

An Inspector Pipeline observes traces across cells, detects patterns, and proposes promotions back into the catalog. The compounding mechanism.

Why It Matters

Most organizations adopting AI in 2026 are stuck in Phase 1 (AI assists individuals; per-individual gains plateau). Some reach Phase 2 (per-team gains; cells operate locally well). Almost none reach Phase 3 (cross-team learning; organizational intelligence compounds).

The 2-to-3 transition is the hard one. It requires architectural investment that doesn't produce per-team gains in the short term. KCC is the specification of what that architecture looks like.

How to Engage

If you...Read
Need a 5-minute architecture briefingArchitectural Overview
Are evaluating adoptionPhase Model and Core Thesis
Are implementingThe full specification
Want the philosophical caseCore Thesis and Cognitive Substrate