Skip to content

engineer@aine:~/book$ open figures/ --all

The visual argument,in 37 diagrams.

Every figure in the book with the caption it ships with. If you want the fastest possible sense of what the book is claiming, scroll this page.

  • 37 figures
  • 11 chapters
  • captions from the manuscript
01

The AI-Native Engineer

chapter page →
figure 1-1
The three pillars of AI-native engineering (intent, constraints, verification) and how each informs the others bidirectionally.
The three pillars of AI-native engineering (intent, constraints, verification) and how each informs the others bidirectionally.
figure 1-2
The eight skills that define AI-native engineering capability.
The eight skills that define AI-native engineering capability.
02

From LLMs to Agents

chapter page →
figure 2-1
The model predicts one token at a time, each time conditioning on everything that came before.
The model predicts one token at a time, each time conditioning on everything that came before.
figure 2-2
The agentic loop.
The agentic loop.
figure 2-3
HITL at artifact boundaries.
HITL at artifact boundaries.
03

Context Engineering Fundamentals

chapter page →
figure 3-1
How context accumulates across a session. The static foundation is established before the first message; dynamic content (like tool calls, tool results, and follow-up messages) fills in as the session progresses.
How context accumulates across a session. The static foundation is established before the first message; dynamic content (like tool calls, tool results, and follow-up messages) fills in as the session progresses.
figure 3-2
Each iteration of the agent loop adds tool results to the context window. The context grows with every action, and managing that growth is part of what context engineering addresses.
Each iteration of the agent loop adds tool results to the context window. The context grows with every action, and managing that growth is part of what context engineering addresses.
04

Model Context Protocol

chapter page →
figure 4-1
The MCP three-party model: The host manages isolated clients, one per server, and each server provides tools, resources, or prompts. The host enforces what each client can access.
The MCP three-party model: The host manages isolated clients, one per server, and each server provides tools, resources, or prompts. The host enforces what each client can access.
figure 4-2
A task-matched configuration can return more than 15% of context budget to actual task content by removing tool descriptions the agent doesn't need.
A task-matched configuration can return more than 15% of context budget to actual task content by removing tool descriptions the agent doesn't need.
05

Spec-Driven Development

chapter page →
figure 5-1
The contrast between vague prompts and spec-driven input to an AI coding agent.
The contrast between vague prompts and spec-driven input to an AI coding agent.
figure 5-2
The Spec-Driven Development loop with human checkpoints.
The Spec-Driven Development loop with human checkpoints.
figure 5-3
From vibe coding to full SDD: a spectrum of approaches ordered by structure and reliability.
From vibe coding to full SDD: a spectrum of approaches ordered by structure and reliability.
figure 5-4
The three maturity levels of Spec-Driven Development: Spec-First, Spec-Anchored, and Spec-as-Source.
The three maturity levels of Spec-Driven Development: Spec-First, Spec-Anchored, and Spec-as-Source.
06

The SDD Workflow

chapter page →
figure 6-1
The six-phase SDD loop and its durable artifacts, with mandatory human checkpoints between Specify/Plan and Verify/Integrate.
The six-phase SDD loop and its durable artifacts, with mandatory human checkpoints between Specify/Plan and Verify/Integrate.
figure 6-2
A task dependency graph showing two parallel prerequisites, a shared downstream task, and two independent tasks that can run in parallel after it.
A task dependency graph showing two parallel prerequisites, a shared downstream task, and two independent tasks that can run in parallel after it.
figure 6-3
The two-tier spec lifecycle model: a strategic layer of living project documents on top, with per-task tactical artifacts below.
The two-tier spec lifecycle model: a strategic layer of living project documents on top, with per-task tactical artifacts below.
figure 6-4
A 2x2 prioritization matrix for brownfield speccing, plotting modules by churn rate against blast radius.
A 2x2 prioritization matrix for brownfield speccing, plotting modules by churn rate against blast radius.
07

SDD Frameworks Compared

chapter page →
figure 7-1
The SDD framework ceremony spectrum, from lightweight conventions on the left to multi-agent enterprise frameworks on the right.
The SDD framework ceremony spectrum, from lightweight conventions on the left to multi-agent enterprise frameworks on the right.
figure 7-2
OpenSpec's two-folder model separates the system's current behavior in `specs/` from proposed modifications in `changes/`, with archived changes merging back into the source-of-truth specs.
OpenSpec's two-folder model separates the system's current behavior in `specs/` from proposed modifications in `changes/`, with archived changes merging back into the source-of-truth specs.
08

Verification and Quality Gates

chapter page →
figure 8-1
The bottleneck has moved. Verification is now the constraint.
The bottleneck has moved. Verification is now the constraint.
figure 8-2
Four automated layers do the mechanical work; the human in the loop steers the pipeline from outside it.
Four automated layers do the mechanical work; the human in the loop steers the pipeline from outside it.
figure 8-3
Two ways to organize LLM review. Start simple; specialize where it pays off.
Two ways to organize LLM review. Start simple; specialize where it pays off.
figure 8-4
The detect-to-fix loop. A production bug reenters the same verification stack as any other change, and leaves behind a test that prevents its return.
The detect-to-fix loop. A production bug reenters the same verification stack as any other change, and leaves behind a test that prevents its return.
figure 8-5
Two mechanisms decide every PR: a deterministic path gate, then the AI scorer for everything else.
Two mechanisms decide every PR: a deterministic path gate, then the AI scorer for everything else.
figure 8-6
One change, end to end. Cheap checks run first and catch the most; expensive human attention is spent last and only where it is needed.
One change, end to end. Cheap checks run first and catch the most; expensive human attention is spent last and only where it is needed.
figure 8-7
The trust ladder. Authority is earned one reversible step at a time, never assumed.
The trust ladder. Authority is earned one reversible step at a time, never assumed.
09

Agent Orchestration Patterns

chapter page →
figure 9-1
Compaction lets the system summarize your context in the dark; a scratchpad puts the same memory in a file you can read, edit, and carry into a fresh window.
Compaction lets the system summarize your context in the dark; a scratchpad puts the same memory in a file you can read, edit, and carry into a fresh window.
figure 9-2
SDD treats the spec as a shared coordination artifact that every phase reads from and writes to, which makes it safe to run different phases on different models.
SDD treats the spec as a shared coordination artifact that every phase reads from and writes to, which makes it safe to run different phases on different models.
figure 9-3
One agent generates, a separate agent critiques, you filter the findings, and the work loops back until it passes a clear bar.
One agent generates, a separate agent critiques, you filter the findings, and the work loops back until it passes a clear bar.
figure 9-4
Three levels of isolation for running agents in parallel: separate context windows, separate worktrees on one machine, and separate cloud machines.
Three levels of isolation for running agents in parallel: separate context windows, separate worktrees on one machine, and separate cloud machines.
figure 9-5
The same subagent building block arranged four ways: orchestrator-worker, sequential pipeline, fan-out/fan-in, and router.
The same subagent building block arranged four ways: orchestrator-worker, sequential pipeline, fan-out/fan-in, and router.
figure 9-6
The dark factory assembles every pattern in this chapter into one self-running line: a spec goes in, agents plan, write, and test in a loop, and a reviewable change comes out.
The dark factory assembles every pattern in this chapter into one self-running line: a spec goes in, agents plan, write, and test in a loop, and a reviewable change comes out.
10

Scaling AI-Native Engineering in Teams

chapter page →
figure 10-1
The AI-native Team Canvas: eight areas, turning "how do we adopt AI?" into a set of decisions a team can answer together.
The AI-native Team Canvas: eight areas, turning "how do we adopt AI?" into a set of decisions a team can answer together.
figure 10-2
AI rules as a tree: a spine from a shared base down through language, framework, and team/project rules, with cross-cutting concerns--testing, observability, infrastructure--branching off to the side rather than stacking on top. Everything is versioned in Git, and different projects share the base while composing their own branches.
AI rules as a tree: a spine from a shared base down through language, framework, and team/project rules, with cross-cutting concerns--testing, observability, infrastructure--branching off to the side rather than stacking on top. Everything is versioned in Git, and different projects share the base while composing their own branches.
figure 10-3
The experiment loop as a flow chart: run an experiment, and if it doesn't find something useful, discard it; if it does, the team discusses and validates it, and only if the team accepts it does it become a standard--then loop back and keep testing new things and assumptions.
The experiment loop as a flow chart: run an experiment, and if it doesn't find something useful, discard it; if it does, the team discusses and validates it, and only if the team accepts it does it become a standard--then loop back and keep testing new things and assumptions.
figure 10-4
Before: Work moved in a straight relay from Product through Design to Engineering to QA. After: A tight loop of three roles working from a shared spec, with an agent fleet handling the mechanical work.
Before: Work moved in a straight relay from Product through Design to Engineering to QA. After: A tight loop of three roles working from a shared spec, with an agent fleet handling the mechanical work.
11

AI-Native in Practice at Cogwheel

chapter page →
figure 11-1
A low-fidelity mockup of Pulse: a catalog of team-health signals, one card per signal, aggregates only.
A low-fidelity mockup of Pulse: a catalog of team-health signals, one card per signal, aggregates only.

Figures are the author's own, drawn for the book and reproduced here for readers. They are the pre-production versions; the published edition carries O'Reilly's redraws.