The 1.0.3 update folds four surfaces shipped at Code w/ Claude 2026 into the chapters where the operator-side discipline meets them. Existing buyers got the new PDF and EPUB free via Lemonsqueezy. The frontier section in Chapter 12 is no longer speculation.
Anthropic ran Code w/ Claude 2026 on May 6. The announcement list was longer than a usual model bump: Claude Managed Agents (multi-agent orchestration, Outcomes, "Dreaming" as a research preview), Claude Code Routines, Agent View, native Code Review used internally across Anthropic teams, CI Auto-Fix, the five-hour-window doubling for Pro/Max/Enterprise, and Claude Platform on AWS for the enterprise side. No new flagship model. The week-after readthrough is: a lot of what was framed as "frontier" in the book's first edition just shipped.
1.0.3 is the integration release. Five chapters touched, ~95 net new lines, no structural changes. The buy URL is the same; existing readers got the update for free in Lemonsqueezy.
Buy Claude Code in Production $39 →
Chapter 5 - Agent View. The new dashboard for managing concurrent Claude Code sessions. Launch with claude agents, drop a foreground session in with /bg, peek at the last response of any agent without attaching, reply inline to ones that need input. The mental shift the chapter spends a section on: the old "don't pile up background subagents" rule was partly coping with the fact that backgrounded work was invisible. Agent View solves the visibility problem. Attention is the new bottleneck. Ten concurrent agents you can see is still ten reports to spot-check.
Chapter 7 - Routines. The productized version of what the book covered as /loop + /schedule. The framing Anthropic used - "wake up to PRs that are ready to merge" - is the shape. The new section keeps the principles that ported (idempotency in scheduled work, cache-window arithmetic, kill switches you've tested) and names what's nicer (outcome-shape contracts borrowed from Managed Agents; the Agent View dashboard as the inbox). The 300-second polling tax is the same tax whether the primitive is named "schedule" or "routine."
Chapter 9 - The five-hour window. Pro, Max, and Enterprise had their usage window doubled. The cost-and-capacity chapter has a new subsection on what this changes (more runway between cap interactions, less arithmetic about pacing a day-long session) and what it doesn't (cost-monitoring discipline gets more important, not less, when the cap moves further out - a bigger window means a bigger possible spend before you notice).
Chapter 11 - Native Code Review and CI Auto-Fix. The teams chapter already covered the agent-as-reviewer and agent-as-formatter patterns with custom GitHub Actions plumbing. 1.0.3 names the native versions Anthropic announced: Code Review (used internally across Anthropic teams) wraps the reviewer pattern with less custom glue; CI Auto-Fix is a tighter-scoped sibling of agent-as-formatter for mechanical-only fixes. Same trust-boundary rule applies in both cases: agents propose, humans approve, an auto-fix that quietly changes runtime behavior is the worst kind of regression. Also one paragraph for the AWS reader on Claude Platform on AWS - it's infrastructure-layer, not a change to how Claude Code works on your laptop.
Chapter 12 - The autonomous-agents section, rewritten. The first edition had three speculative bullets here: long-horizon autonomous agents, multi-agent collaboration, self-modifying overlays. May 6 shipped meaningful pieces of all three. The new section names Managed Agents (multi-agent orchestration + Outcomes + Dreaming), Routines, and Agent View as what landed against those bullets - and what didn't (no new flagship model, which may mean the model-bump treadmill the chapter assumed stretches as Anthropic shifts attention to the agentic-product surface). The principle the chapter rests on is unchanged: the disciplines port. The procedural rules from Chapters 5 through 11 map directly onto the named products. The failure-mode catalogue for the new shapes will fill in over the next twelve months.
Updates are part of what the book sells. Lemonsqueezy emailed every existing buyer the new PDF and EPUB on release; the download links in your receipt point at the 1.0.3 build. If your inbox didn't get it, the order page at app.lemonsqueezy.com/my-orders has the latest files.
The companion repo at YawLabs/claude-code-in-production-companion is unchanged in this release - the exercises and worked answers still match the chapter targets. The next round of companion updates will line up with the failure-mode catalogue work mentioned above.
The Anthropic surface area is moving faster than the book can reasonably track at the level of "every command and flag." That's fine - the book was designed for it. The four-part structure (foundations, workflow, reliability, beyond solo), the seven hazards, the verification posture, the cache-window arithmetic, the scope-discipline rules - those don't move when a new product surface ships. What moves are the surfaces the disciplines operate over, and 1.0.3 is the maintenance pass for that.
Watch the changelog. The next bump will be either the next Anthropic event with operator-relevant surface (likely the next quarterly window) or the moment failure modes for the new shapes start clustering enough to be worth their own section. Whichever comes first.
Twelve chapters of operator-side discipline. Now 1.0.3.
PDF + EPUB. Free updates as the surface moves. $39 one-time, secure checkout.
Published by Yaw Labs.