Guided Intelligence
Read paper
§2 · Core Concepts

Humans set direction. AI does the work.

GI rests on four ideas that shape every role, workflow, and safety mechanism in the system.

§2.1

Human-in-the-Loop as a First Principle

Humans no longer execute most of the work. They guide it, constrain it, and validate it. HIL in GI isn't "humans approving AI output." It's a structured responsibility model.

  • Planners set intent, constraints, and architectural boundaries.
  • Builders supervise AI execution and keep it faithful to those constraints.
  • Reviewers perform semantic approval and protect domain invariants.

The goal isn't to remove humans. It's to separate humans from work that no longer requires them.

§2.2

From Mechanical to Semantic Work

Mechanical work moves to AI: coding, scaffolding, refactors, formatting. What is left is the semantic layer: what should be built, why it matters, and whether it preserves system integrity.

  • Mechanical cheap. Deterministic. Executed by AI.
  • Semantic expensive. Judgment-driven. Owned by humans.

The profile of engineering work changes: fewer executors, more stewards.

§2.3

Continuous Flow vs. Batching

Batching (sprints, release trains, QA cycles, PR queues) stabilizes human-paced work by delaying it. When AI executes instantly, that same batching becomes the main source of latency.

  • No sprint boundaries work moves to the next step as soon as the current one completes.
  • No release windows changes deploy continuously through rings.
  • No PR queues one Reviewer is the semantic gate.

The pipeline assumes work should be moving at all times unless halted for safety.

§2.4

The 10× Hypothesis

AI writes code 100× faster. The 10× system-level multiplier comes from removing the human overhead around that code, not from typing faster.

  • Limiting reagent shifts from engineering labor hours to semantic clarity.
  • Direction "10×" is a target architecture for velocity, not a fixed performance number.

GI creates the conditions for order-of-magnitude improvement as AI continues to advance.

§1.1 · Why Agile breaks under AI

Same rituals. Opposite effect.

Agile rituals stabilized human execution. Under AI acceleration, these stabilizers reverse polarity.

Agile
Optimized for human execution
Guided Intelligence
Optimized for AI execution + human semantics
Batch-driven sprint cycles
Continuous unbatched flow
Standups, grooming, planning meetings
Semantic clarity via Execution Plans
Manual user-story decomposition
AI generates scaffolding from intent
PR queues stalled on review
Single Reviewer as semantic gate
End-of-sprint QA batches
PAT validates the running system
Scales with headcount
Scales with AI capability
§1.2 · The new constraint

"If AI does mechanical work orders of magnitude faster, engineering productivity becomes a function of human semantic clarity, not human labor volume."