Skip to main content

What's New

Track meaningful updates to the book. This page is refreshed once a day from the latest changes to the book's content.

Latest Updates​

Updates are grouped by day and written automatically from the day's content changes, each linking to the lesson that changed.

2026-09-11​

2026-09-09​

2026-09-08​

  • Added a new revenue section to The Agent Factory Thesis, showing AI sales crossing the hardware depreciation line and sharpening the bubble question. (The Agent Factory Thesis)
  • Expanded Harness Engineering with new guidance on context surfaces, agent-experience design, hooks, and typed output, plus a live hook walkthrough. (Harness Engineering: A Crash Course)

2026-09-06​

2026-09-05​

2026-09-04​

  • Added a new teaching aid slide deck to AI Fluency so the 4Ds lesson now has a presentation readers can open alongside it. (AI Fluency: A Crash Course)
  • Replaced the AI Prompting 2026 teaching deck with a refreshed presentation for the current version of the crash course. (AI Prompting in 2026: A Crash Course)
  • Restored the Workflow Design & Diagnosis flashcards to the full 30-card set so the practice deck matches the course again. (Workflow Design & Diagnosis)

2026-09-03​

2026-09-02​

  • Reworked the certifications path around PCAO-F first and PCAR-F second, with clearer sample timing, exam guides, and what the Panaversity and Anthropic credentials each mean. (Certifications: Proof You Can Carry In)
  • Expanded AI Fluency, AI Prompting, and Claude and ChatGPT 101 to line up better with the CCAO-F guide and the way the Foundations path now flows. (AI Fluency: A Crash Course)
  • Sharpened Workflow Design & Diagnosis so it now walks through step ownership, requirements extraction, diagnosis by symptom timing, and fixes that stick. (Workflow Design & Diagnosis)
  • Tightened Governance, Risk & Responsible Use with a clearer four-question decision flow for when AI can do the work, what data can go in, what tools can turn on, and when to escalate. (Governance, Risk & Responsible Use: A Crash Course)
  • Updated General Agents on the Web to explain cloud versus local execution, desktop bridges, and the six-part harness model behind the major web agents. (General Agents on the Web: A Crash Course)

2026-08-31​

  • Added seven Claude platform crash courses, covering the Messages API loop written by hand, structured extraction pipelines, the Claude Agent SDK, Claude Code for teams and as a CI worker, routines, and managed agents. (Build AI Agents with the Claude Agent SDK)
  • Added hands-on projects to the end of each of the seven Claude courses, mapped to the preparation exercises in the official Architect Foundations exam guide, each with an observable finish line and a step that breaks the thing on purpose. (The Loop by Hand: The Anthropic Messages API)
  • Added a full-length practice exam for the Architect Foundations path, reachable from the certifications page, with a worked answer key and a per-domain score breakdown. (Certifications: Proof You Can Carry In)
  • Added flashcard decks to all seven Claude courses, and rebuilt every quiz answer set so the correct option is no longer the longest one on the page. (Structured Extraction Pipelines)
  • Turned the practice exam into a playable simulator, with a 120-minute timer, one question at a time, and a score report that breaks your result down by exam domain and replays every question with the reasoning behind the right answer. (CCAR-F Practice Exam)
  • Fixed 25 headings in the Spec-Driven Development course that never created a link target, so links pointing into those sections now land where they should. (Spec-Driven Development: A Crash Course)

2026-08-30​

  • Aligned five getting-started crash courses to the Associate Foundations objectives, adding a real-work move to each rather than more explanation. (AI Fluency: A Crash Course)

2026-08-28​

2026-08-26​

2026-08-25​

2026-08-23​

2026-08-22​

2026-08-21​

2026-08-19​

  • Clarified in Graph Engineering that the phrase carries more than one meaning, and named which two the course teaches: the memory graph your loops share, and the governance graph wiring the loops to each other. (Graph Engineering: A Crash Course)

2026-08-10​

  • Added The Ecosystem Concept as the opening page of the ecosystem section, with its own flashcard deck. (The Ecosystem Concept)

2026-08-06​

  • Rewrote The Four Layers as the map for the rest of the section, opening on the forty-minute run that cost fifty dollars and produced nothing, and adding a diagram of the six-node graph. (The Four Layers: Prompt, Context, Harness, Loop)

2026-08-05​

  • Added How to Sell in the Agentic AI Era, on why proof replaces the pitch, and closed the roadmap with a pointer to it. (How to Sell in the Agentic AI Era)
  • Added Building the Context Layer, a crash course on going from one worker's store to a shared one. (Building the Context Layer)
  • Editorial pass on the preface, the about page, How to Learn from This Book, The Roles This Book Trains, Why AI Is Non-Negotiable, and The Agent Is the Operating Layer. (Preface: The Right Side of the Line)

2026-08-04​

2026-08-03​

  • Refined the Zia Tutor AI onboarding and completed its change list. (Zia Tutor AI)

2026-08-02​

  • Moved Zia Tutor AI to its own address, with the old link redirecting. (Zia Tutor AI)

2026-08-01​

2026-07-31​

  • Retired the just-in-time chapter tier, so Parts 0 through 7 and the Project Labs are no longer part of the live book and their old links now land on the crash courses. (Getting Started: Crash Courses)
  • Added The System of Context as the closing page of the ecosystem section, on the layer above a portfolio of governed records: a System of Record decides what is true, a System of Context decides what arrives. (The System of Context: Connecting the Records to Real Work)
  • Added Made, Not Generated, a new crash course on building websites with taste files, screenshots, and publish-then-review loops. (Made, Not Generated: A Website Design Crash Course)
  • Reworked the vertical-picking path with a 2026 Big Four reality check, and tied the SoR and thesis chapters to the same business-transformation warning. (Choosing Your Vertical)

2026-07-30​

  • Expanded The Roles This Book Trains with Kevin Bai's FDE 101, the do-I-need-an-FDE test, and the dev-shop warning, with the same shared-primitives idea echoed in the FDE AF Model. (The Roles This Book Trains)

2026-07-29​

  • Grew Local AI into Open Source LLMs: Your Laptop, Your Server/Cluster, the Cloud. The old local-model course is now Part 1, and two new parts follow it: serving the same model to fifty people at once with vLLM and drawing your own two throughput curves, then driving frontier open models like Kimi K3 and DeepSeek V4 Pro through OpenRouter and routing all three tiers from one config. The old link redirects here. (Open Source LLMs: Your Laptop, Your Server/Cluster, the Cloud)

2026-07-28​

2026-07-27​

  • Added What You Carry In, the book's third argument, on what a practitioner can still own when the model is rented and the method is given away free, written with plain-words boxes and a fifteen-term glossary for readers whose first language is not English. (What You Carry In: The Ownership Argument)
  • Wired the new ownership argument into the pages that frame it, so the thesis, the operating layer, and this page now read as one set of three. (The Agent Is the Operating Layer)
  • Answered the worry sitting under Code You Never Write, using the author of Clean Code, who stopped reading the code his agents write only after he had built and measured the checks that let him. (Code You Never Write: A Crash Course)
  • Moved graph engineering out of Loop Engineering so it can stand as its own course, and pointed the two at each other. (Loop Engineering: A Crash Course)

2026-07-26​

  • Added the missing half of the AGENTS.md method to the agentic coding course: delete a rule once the model no longer needs it, worked through Anthropic's own July 2026 cut of more than eighty percent of the Claude Code system prompt. (Claude Code and OpenCode: A Crash Course)
  • Added memory to the prompting course as the sixth layer of the context stack, and explained why a tool that remembers you across chats does not make the model any less stateless. (AI Prompting in 2026: A Crash Course)
  • Added a seventh question to the Python review loop, on why messy code now costs tokens and retries rather than only your own patience six months later. (Python in the AI Era: A Crash Course)
  • Named the honest limit of specs in the spec-driven course: the spec is the source of truth for what the system does, and the code is still the only evidence of how well it is holding together. (Spec-Driven Development: A Crash Course)

2026-07-25​

  • Added Designing the Vertical System of Record, the method for deciding what belongs inside a domain record instead of collecting a profession's current workflows and calling that a corpus, with two worked appendices, six templates, a quiz, and a flashcard deck. (Designing the Vertical System of Record from First Principles)
  • Added How to Learn from This Book, which lays out the learning system, the nine-stage learning loop, and the seven-step routine for working through a crash course. (How to Learn from This Book)
  • Rebuilt The Roles This Book Trains around the two Systems of Record a vendor-neutral engineer carries, and gave it a teaching aid, a flashcard deck, and a fifty-question quiz. (The Roles This Book Trains)
  • Threaded the thin slice through the ecosystem pages, so choosing a vertical, designing its record, and the factory model all describe the same first build. (The Forward Deployed Engineer Agent Factory Model)
  • Surfaced the teaching aid deck above the glossary in General Agents on the Web. (General Agents on the Web: A Crash Course)

2026-07-24​

2026-07-23​

  • Added Choosing Your Vertical, a guide to picking the profession your domain startup serves, with the screen, the eight tests, the four launch gates, and its own quiz and flashcards. (Choosing Your Vertical)
  • Added a teaching aid, a flashcard deck, and a quiz to The Forward Deployed Engineer Agent Factory Model. (The Forward Deployed Engineer Agent Factory Model)

2026-07-22​

2026-07-21​

2026-07-20​

2026-07-19​

  • Added Local AI: Agentic Coding on a Model You Run, a course on running a model on your own machine for free and offline, then pointing a coding agent at it. (Local AI: Agentic Coding on a Model You Run)
  • Expanded What AI Actually Is with new explanations of why the model is frozen, how large a context window really is, how chat history is replayed on every turn, and where Skills fit, plus a bigger flashcard deck. (What AI Actually Is: A Crash Course)
  • Corrected nine errors that were taught as fact and fixed quiz and companion drift across two dozen Part 4 Python chapters. (Part 4: Programming in the AI Era)
  • Defined the funnel pattern and added a worked quality-checklist example. (The Complete Feedback Loop)

2026-07-18​

2026-07-17​

  • Added Trusting the Checker, a crash course on how to know whether the thing checking your agent's work can be trusted. (Trusting the Checker: An Evals Crash Course)
  • Added Harness Engineering, a crash course on the five verbs that shape what an agent is allowed to do: constrain, inform, verify, correct, and escalate. (Harness Engineering: A Crash Course)
  • Gave Harness Engineering eight interactive simulations, thirteen short drills with a prompt to paste and something you can watch happen, and a teaching slide deck. (Harness Engineering: A Crash Course)
  • Put runnable projects inside Loop Engineering itself, so you watch a loop fire where it is taught rather than seven hundred lines later. (Loop Engineering: A Crash Course)

2026-07-16​

  • Added six interactive simulations covering the in-session loop, goals, routines, the event-driven doorbell, the save-point spine, and one flagship walkthrough of a whole morning. (Loop Engineering: A Crash Course)
  • Added a section on dreaming, the separate loop that consolidates an agent's memory, along with its failure modes and new flashcards. (Loop Engineering: A Crash Course)

2026-07-15​

  • Clarified the Loop Engineering crash course with plainer guidance on background sessions, timers, scheduling, retries, and stopping loops, plus a cleaner orientation for the event-driven section. (Loop Engineering: A Crash Course)

2026-07-14​

2026-07-13​

2026-07-12​

2026-07-11​

2026-07-10​

2026-07-09​

2026-07-08​

  • Expanded Loop Engineering with a clearer four-layer mental model, four more practice projects, and three routine drills in the appendix. (Loop Engineering: A Crash Course)

2026-07-07​

  • Reframed the Agentic Coding crash course around trust-by-checking, added a verification loop section and a new maturity ladder, and expanded the quiz. (Claude Code and OpenCode: A Crash Course)
  • Clarified Loop Engineering by separating the inner loop from the outer loop and spelling out how the layers fit together. (Loop Engineering: A Crash Course)
  • Tightened The Agent Factory Thesis with first-party Claude Code evidence and a clearer line between invariants and implementation. (The Agent Factory Thesis)

2026-07-06​

2026-07-05​

  • Reworked the preface to frame the book around the line between people who build AI workers and people who get displaced by them. (Preface: The Right Side of the Line)
  • Expanded the loop engineering crash course with a clearer outer-loop vs inner-loop model and new diagrams for the six-part loop shape. (Loop Engineering: A Crash Course)

2026-07-04​

  • Expanded The Roles This Book Trains into a full map of the FDE market, including salary ranges, freelance work, interview prep, and a five-archetype view of the job. (The Roles This Book Trains)
  • Tied The Agent Factory Thesis more directly to the Forward Deployed Engineer path and the AI-Native Company Architect. (The Agent Factory Thesis)

2026-07-03​

2026-07-02​

2026-07-01​

2026-06-30​

2026-06-29​

2026-06-28​

2026-06-26​

  • Fixed the Getting Started course map so the engineer path now points to Course 17 for Build AI Agents and keeps the Digital FTE sequence aligned. (Getting Started: Crash Courses)

2026-06-25​

  • Reframed Personal Agent Harnesses as the shared step where readers build an owned persistent agent, with clearer OpenClaw and Hermes choices. (Personal Agent Harnesses)
  • Added a new Mode 1 path with a task-diagnosis lesson and a bridge from one-off solves to workers. (Mode 1 — Problem-Solving)
  • Added section overview pages for the General Agents and Mode 2 Manufacturing tracks, so each track now opens with a guided map. (General Agents, Mode 2 — Manufacturing)

About This Page​

This book is continuously updated. Instead of version numbers, we track changes here so you can see what's new since your last visit.

  • Each entry links to the specific lesson that changed
  • A short digest is written once a day from the latest content changes on main
  • Only changes to book content (not infrastructure) appear here

Tip: Bookmark this page to quickly check for updates.