Get your engineering team to AI-native — without the chaos.

The Playbook is the curriculum

The AI-Native Engineering Playbook: Crawl, Walk, Run, Fly is the framework I’ve been using with engineering teams for the past couple of years. Four stages, in order, that you can’t skip.

• Crawl — a traditional competent team that ships software, maybe ChatGPT on the side. Most engineering orgs today.
• Walk — every engineer has had their “letting an agent change my files” moment. The capability jump landed. In scope for this bootcamp.
• Run — everything as markdown in the repo. Jira killed. Meetings re-engineered around transcripts. Claude owns git. Cross-team sync meetings disappear. In scope for this bootcamp.
• Fly — every engineer running 4–6+ parallel Claude Code sessions on their weekly objectives. The elite practitioner tier — AgentHerder Bootcamp sits on top of this one.

This bootcamp takes your team from Crawl through Run. The Fly tier — the elite-practitioner massive-parallelism practice — is what AgentHerder Bootcamp is for, and it sits on top of this one. You graduate this and the next step is clear.

One ritual installed from Week 1 and run every week of the program: Friday demo + weekly planning. Each engineer demos their week’s work, then plans next week, in front of their team. About an hour. By the end of the bootcamp it’s reflexive, not an exercise — and it’s the cadence the running-stage practice runs on.

Why six weeks. Why your real repo. Why team-based. Why the ritual.

Four pedagogical commitments that materially change whether the practice sticks:

1. Spacing, not bingeing. The capability jump from “I tried Claude Code” to “I use Claude Code by default” takes weeks, not a workshop weekend. Spreading it across six weeks — with real work in between — is how the habit forms. A one-day training doesn’t.
2. Your repo, not a sandbox. The whole point of the practice is that AI is only useful when it has your specific context. Every assignment runs against your actual codebase, with your actual conventions and constraints. Generic tutorials don’t transfer; this does.
3. Team practice, not individual heroics. A bootcamp for one engineer is a bootcamp for someone who’ll burn out being the only one who works this way. A team that goes through together has shared vocabulary, shared skill libraries, and shared rituals by week six — which is what compounding actually requires.
4. The Friday-demo + weekly-planning ritual, every week. One hour. Each engineer demos the week’s work and plans next week in front of the team. Installed at end of Week 1 and run every week of the bootcamp. By graduation it’s reflexive, not an exercise — and it’s the cadence the running-stage practice actually runs on. Without this ritual, the rest of the curriculum doesn’t compound.

The guarantee

Six weeks. AI-native practice in your team’s real workflow. Or full refund.

Show up. Do the assignments on your real repo. Run the Friday demo + weekly planning ritual every week of the program. If by graduation your team can’t demonstrate the six concrete artefacts — the repo-as-context migration, the new-operating-routines kit (browser automation + transcription pipeline + deploy-to-prod skill), the 10×-everything function audit, ultra-fast iteration in your real workflow, calibrated quality + a mock/replica unblocking your team from a slower upstream, and a graduation demo of work you couldn’t have shipped six weeks ago — refund, no questions.

This is outcome-bound, not satisfaction-bound. Not “if you didn’t enjoy the videos.” If the practice doesn’t take in your team, your money comes back. The bar is the same one I hold myself to in my own daily practice.

FAQ

What does the team need to commit to for this to work?

Two things, both hard prerequisites. (1) Real production code throughout the six weeks — the participant’s everyday work, or a new feature / module being kicked off, or a new initiative using the bootcamp as launch ramp. Not side projects, not toy repos, not “I’ll apply this after the bootcamp ends.” The curriculum is mostly behaviour change between sessions, and behaviour change without real stakes doesn’t take. (2) The Friday demo + weekly planning ritual, every week, ~1 hour. Both are non-negotiable.

Will my team need to install something painful on corporate laptops?

Mostly: Claude Code (Claude.ai subscription or Anthropic API key), the gh CLI for the git week, and a local Whisper install for the transcription week. Nothing exotic. Most corporate environments allow all of this; some block outbound traffic to Anthropic’s API or restrict global npm installs — worth checking with IT before enrolling a whole team. If a personal-laptop setup is needed for some of the assignments, that’s fine too.

Mac, Linux, or Windows?

Claude Code runs on all three. The transcription pipeline (Whisper) is happiest on macOS or Linux; Windows works under WSL2. Native Windows is possible but harder. If most of your team is on Windows without WSL, message me before enrolling and we’ll talk through it.

Is this for the senior IC trying to roll AI out, or for the team as a whole?

Both work. The individual tier exists for the IC who’s been told to “figure this out for the team” — they do the bootcamp, then carry the practice back. The team tier exists for engineering leaders who want the whole squad through together. The compounding-returns case is much stronger with the team tier, because the practice is by design something a team does together.

How async is it?

Fully async. No live calls. No scheduled cohorts. No mandatory Zooms. You can start any time. Cohorts — everyone who starts in the same month — get a private channel for shared questions, but the cadence is “reply when you reply,” not synchronous.

How is this different from Claude Code documentation, YouTube tutorials, or the Anthropic Architect cert?

Documentation and tutorials teach you tools. The Architect cert (free for Claude Partner Network members, since March 2026) validates your knowledge of Claude features. Neither teaches a team how to adopt the practice. The bootcamp is the curriculum I wish had existed when I was rolling this out at F-Secure and DNV Cyber: how to sequence the changes, how to handle the team-wide resistance points, what to kill and what to keep, what the rituals look like on the other side. It’s a complement to the Architect cert, not a substitute — some teams will want both.

Does the team need to be on Claude specifically? What about Cursor / Copilot / Codex?

The week-one “Crawl” material is tool-agnostic — Cursor, Codex, Claude Code, all give you the “agent changes my files” moment. Weeks 2–6 lean on Claude Code specifically because that’s where the skills library / mission file / CLAUDE.md / sub-agents stack is most mature today. If your team is committed to a different stack, message me — the patterns mostly translate, but I won’t pretend they’re identical.

Will this work if my team is mostly senior engineers?

Yes — arguably better. Senior engineers already have the judgement to evaluate when Claude is “declaring victory” on work that isn’t actually done. They also have the political capital to land the cross-team and process-rethink changes in weeks 4–6. The bootcamp accelerates their adoption; it doesn’t replace their judgement.

What if the team finishes week six and wants more?

The next step is the Fly tier — the elite-practitioner massive-parallelism practice. That’s AgentHerder Bootcamp, which sits on top of this one. Graduates of this bootcamp who want to push further (running 8+ parallel Claude Code sessions on their weekly objectives) are exactly the AgentHerder audience.

Is there a certificate?

A team retrospective deliverable that documents what the team built across the six weeks, signed off by me. Not a multiple-choice exam — the deliverable IS the credential, and it’s something the team can show internally to justify the spend. The standalone AgentHerder Certified credential (parallelism-specific, real-repo deliverable) is separate and lives under AgentHerder — details when that exam launches.

When does the next intake start?

First of every month. Reserve a spot below; I’ll email when the next intake opens.

Honest failure modes — when does the bootcamp not work?

Three cases I’ve seen. (1) Management blocks process change. If your manager won’t let you drop Jira / sprint ceremonies / handcraft-code conventions, the capability gets installed but the throughput stays capped. The team tier exists partly because shipping a whole squad together makes this conversation easier. (2) Support functions don’t scale with engineering velocity. 10× more PRs without 10× more QA / monitoring / security = 10× more bugs. Week 3 (the 10×-everything chapter) addresses this directly, but those surrounding functions need to be in the conversation. (3) One graduate inside a non-AI-native team gets isolated. The 5-seat package exists for this reason — install across one team, not one person. Year-1 outcomes range from 1.5× (heavy org friction) to 3× (organizational autonomy). Honest expectation.

What if I’m looking for AI in my personal life, not my engineering team?

That’s a different program — same six-week structure, different curriculum, designed for individuals (ChatGPT / Claude users adopting AI as a steady presence across life and work). See Adopt AI Bootcamp. Both bootcamps are under the same Augmented Mind umbrella — same standard of quality, different audience.