A 6-week bootcamp for engineers who are done coding and ready to herd. From 1–3 parallel Claude Code sessions to 10+ sustained on real work. From individual productivity to fleet-scale.
The course doesn't start with tools. It starts with the identity transition — because the leap from "a few agents I supervise carefully" to "a fleet of agents I orchestrate" feels qualitatively different, not just quantitatively. The tooling is teachable in a day. The identity work is what fills six weeks.
Reserve your seat → $2,495 individual · $9,995 for a 5-seat team · 6 weeks · Outcome-guaranteed · Cert exam bundled
Want to read the curriculum first? The six weeks are also available free + self-paced as the AgentHerder Course.
You already operate at running stage. You don't write code anymore. Your repo is the context. Your meetings exist for the transcripts. Your team ships PRs as narratives. You've crossed the rubicon that 95% of engineers haven't.
But you've also hit the ceiling running stage imposes. Most engineers cap at 2–3 sustained sessions. Past that, context loss and brain-fry take over and the agents start drifting while you're not looking. And every minute spent clicking through dashboards, drafting routine Slack replies, or chasing partner-portal MFA flows is a minute that doesn't go to the work that actually compounds — context clarification, creative direction, the meetings worth transcribing.
What I've watched, running my own fleet daily and coaching engineers on theirs: ordinary engineers stall at 3 sessions, other engineers accelerate past 8 with minimal coaching. The difference isn't IQ. It's tolerance for the felt experience of flying-stage practice — letting go of context-holding, trusting the structure, the emotional discipline of not-checking. The hiring filter for AI-native teams is temperament as much as résumé, and the same is true for who can sustain a fleet.
That's not a tooling problem. That's an identity-and-coordination problem — and there's no curriculum for it yet.
There's no book. No conference talk. No course. The handful of engineers running 10+ parallel agents today figured it out by running into walls and writing their own tooling. The Bootcamp is the path they didn't have — assembled, sequenced, and tested against real work.
This bootcamp is heavy because the work is heavy. The technique is teachable in a day; the rest is identity work, emotional tolerance, and the coordination patterns that turn 10+ parallel streams from chaos into flow. Most weeks are group + personal coaching, not lecture: framework material to anchor each week, then time spent on what's working, what's blocking, and working through participants' specific issues.
The 2–3 → 10–15 leap. The "almost never say no" reflex — at running stage you might still say "we'll get to that next sprint"; at flying stage you open a new tab and have a rough version this afternoon. Browser automation as the coordination layer: every web interaction that's pure coordination overhead — SaaS dashboards, internal tools, partner portals, GitHub clicking — routes through Claude. Comms automation for the routine surface: status updates, scheduling, inbox triage, routine Slack/Teams/email replies, drafted-and-approved. Your remaining attention goes to the human work that compounds — context clarification with teammates, creative direction, the meetings worth transcribing.
Practical artifact: a 1-page personal manifesto on what changes about how you work. Submit to cohort, read, react.
The mechanical layer. Necessary but not sufficient. cctabs setup: per-tab session management, renaming, fork, restore-after-reboot, multi-window organization. Worktrees: when to branch, when to consolidate, when to throw one away. The foreground-vs-background stream model: at any moment 1–3 streams in the foreground (running dev server, doing live QA, browser-automating against staging), the rest backgrounded. Anthropic quota management: how to be efficient at high parallelism, what to do when quota windows hit, local-model fallback options (Ollama, Kimi, Qwen) via cctabs backends.
Practical artifact: operate 6+ parallel sessions for one full working day, screen-recorded for self-review.
The integration thesis: give your agents more and more integration points so all the small things across every parallel stream cancel out. Each new integration removes a class of coordination friction. The fleet's throughput is determined by the surface area of what the agent can reach without you — which is the inverse of how much of your time is left over for the human work that matters. Browser automation as a deep discipline. Comms automation as a deep discipline.
And the "fully autonomous agents" question — every cohort asks it. The honest answer: you need supervision and human guidance — that's why parallelism is important. Parallelism IS the human operator. We're not here to replace ourselves with agents that don't need us. The value-add is YOU — the domain expertise, the seeing where the agent should be, the guiding. This bootcamp trains you to be the irreplaceable orchestrator of a large agent fleet, not to set up agents that run without you.
Practical artifact: identify one integration your current setup is missing (browser-automation surface, comms-automation surface, or new tool surface) and ship it.
The operational layer that turns 6+ parallel sessions from chaos into flow. The Team Ops sweep skill: a periodic agent that scans across all your tabs, identifies which need attention, surfaces PRs needing review, drafts your responses — the meta-agent that keeps the rest productive. Cross-stream context sharing: when information from one tab needs to flow to another without you being the human router. PR-nudging automation: agents that watch your PRs, nudge reviewers, retry CI, handle draft toggles. Friday demo + weekly planning at flying-stage — different from running stage because of the volume of work-streams to demo and re-plan.
Practical artifact: Team Ops sweep skill installed and running on your own work. You demo it.
The identity work specific to flying stage. Letting go — externalizing context into each tab's working memory instead of holding it all in your head. "It's scary for lots of people and feels awkward and feels like losing control." Brain-fry mitigation — what actually causes it, what doesn't, what to do when it happens; most "I can't do parallelism" stories trace to brain-fry, not capability ceilings. Trusting the structure — when the substrate is right, the agents work even when it feels like nothing is; the discipline of not-checking, of letting streams cook. The political layer — selling parallelism to a skeptical manager, defending your throughput in code review, explaining your work to colleagues who think you're cheating.
Practical artifact: a personal-experience post (cohort-only, not for publication) on your hardest moment in the bootcamp so far, what almost made you quit, what you did instead.
Run ≥12 parallel sessions for one full working day, three days in a row, while shipping real work. AgentHerder Certification exam — timed, observed by a Principal AgentHerder, graded on stream count sustained, output quality across streams, coordination hygiene, decision-making about when to bring streams to foreground, and recovery from quota / context / brain-fry incidents. Graduation cohort call.
Practical artifact: the cert. Or the not-yet (one free re-take within 6 months).
Six weeks of identity work disguised as skills training. Mostly group + personal coaching; lectures only where the material is purely mechanical.
This bootcamp's curriculum is 20% technique and 80% identity. The technique can be taught in a day. Why does the rest take six weeks?
The leap is qualitative, not quantitative. Going from 1–3 to 10+ parallel sessions isn't "the same thing, more of it." It changes how you hold context, how you make decisions, how you experience a working day. Identity transitions don't compress into a weekend.
Real-repo difficulty beats safe practice. Toy repos that always work don't build operators. Your real codebase, with its real merge conflicts, its real CLAUDE.md weirdness, its real production constraints, is what pushes the practice through the plateau where most people stall.
Spaced beats crammed. A weekend bootcamp is the worst possible architecture for skill durability at this scale. Six weeks of one-pattern-per-week is roughly the right size for the practice to embed in your daily work, not stay in a Notion doc you never reopen.
Generation beats showing. Each week's curriculum has you attempting the work first, then walking through how I'd approach it. Trying — and getting stuck — before being shown the answer is what makes the lesson stick. Skipping straight to the answer feels efficient. It isn't.
The emotional work needs time. Week 5 is named "the emotional spine" for a reason. Letting go of context-holding, learning to trust the structure, navigating the political layer with colleagues who think you're cheating — these aren't taught in lectures. They're worked through, in cohort.
The 6-week curriculum is the core. Most of the value is the coaching and the cohort. The assets below are the operational substrate.
Bonus #1
Command patterns and recipes for the daily fleet workflow. How to fork a stuck agent. How to consolidate a five-agent investigation across mission files. How to keep mission files clean. How to recover when an agent goes off-course without losing its work. The unwritten parts.
Bonus #2
The meta-agent that keeps the rest of your fleet productive. Scans across all your tabs, identifies which need attention, surfaces PRs needing review, drafts your responses. The skill you install in Week 4 and never turn off.
Bonus #3
The briefing format that turns "go fix this" into agent-executable work. The template structure and the conventions for the parts that actually matter — what the agent needs to know, what to leave out, what makes a mission file re-runnable when an agent gets stuck.
Bonus #4
The dashboards, partner portals, MFA flows, calendar tools, status-update threads, and inbox-triage patterns I've automated personally. How to handle auth, broken JS, slow pages, the "Claude says it can't, push back" routine. The integration surface that turns coordination overhead into agent work.
Bonus #5
What AI-native engineering adoption actually looks like inside cybersecurity firms. Drawn from coaching engagements at F-Secure and DNV Cyber. What worked. What didn't. What you'd want to know before pitching it inside your org — and the responses to expect from security, procurement, and senior management.
Included
Three months of the Club included with the Bootcamp. Monthly drops of new skills, mission file patterns, and case studies. After 90 days it's optional and continues at $37/mo (Community) or $97/mo (Engineering). Club isn't gating — the Bootcamp stands on its own.
And the AgentHerder Certification exam — bundled with the Bootcamp. Timed, observed by a Principal AgentHerder, graded on practice not knowledge. (You can also take the cert without the Bootcamp; you can take the Bootcamp without sitting the cert. Most people will do both.)
Six weeks. 10+ parallel sessions sustained on real work. Or full refund.
Show up. Do the assignments on your real repo. Sit the cert exam. If at the end of the program you can't sustain 10 or more parallel Claude Code sessions productively on real work — refund, no questions.
This is outcome-bound, not satisfaction-bound. Not "if you didn't enjoy the videos." If the practice doesn't take, your money is back. The bar is the same one I hold myself to running 12–15 sessions daily.
Launch-cohort pricing. May rise once the first cohort completes.
Individual
$2,495 one-time
Team — 5 seats
$9,995 one-time
After the included 90 days, AgentHerder Club continues as an optional recurring layer:
Club is optional. The Bootcamp doesn't gate on it.
Hard prerequisites: you've shipped real work using Claude Code regularly for at least one month, you currently operate 1–3 parallel sessions productively, and you have a real codebase + real work to operate against during the bootcamp. The application asks six open-text questions to screen for temperament and self-awareness (calibration on what you've sustained, tolerance for waiting/context-shifting, experience externalizing working memory). Fred reads every one personally for the v1 cohorts.
No — and we'll redirect you to the Augmented Mind Bootcamp, which is the on-ramp. AM takes engineering teams from walking to running stage. AgentHerder takes running-stage engineers to flying. Trying AgentHerder before AM doesn't compress the path — it just makes the AgentHerder weeks feel disorienting because the running-stage substrate isn't in place.
Three. (1) Brain-fry in Week 2. When sustained 6-stream practice hits, about 1 in 4 candidates needs to slow down or drop. We screen for this in the application but the screen isn't perfect. (2) Org friction caps your stream count regardless of skill. Solo IC in a 100-person traditional org may install the practice and have nowhere to deploy it. (3) "Letting go" doesn't click. Engineers with deep need-for-control patterns can install all the mechanics and still struggle with the emotional spine in Week 5. The bootcamp gives you the chance to find out which (if any) is yours — it doesn't guarantee you'll like flying.
Mostly ICs. Engineering managers can take it to understand the practice from the inside (which is the prerequisite to evaluating engineers who claim they're already doing it) — but the bootcamp is built around an individual operator running their own fleet, not around managing other operators.
cctabs currently runs in Wave Terminal and Tabby on macOS and Linux. Windows works under WSL2; native Windows is on the cctabs roadmap. If you're on Windows today and don't have WSL, the bootcamp will be harder — possible, but harder. Message me before enrolling if that's your setup.
Mostly yes, with caveats. cctabs is npm install -g. Claude Code needs either a Claude.ai subscription or an Anthropic API key. Some enterprise environments block outbound traffic to Anthropic's API or restrict global npm installs — check with your IT team before enrolling. If you can't get Claude Code running on your work laptop, you can do the bootcamp on a personal machine.
This bootcamp is explicitly not training people to set up agents that run without supervision. The premise is the opposite: parallelism IS the human operator. We're not here to replace ourselves. The value-add is you — the domain expertise, the seeing where the agent should be, the guiding. The bootcamp trains you to be the irreplaceable orchestrator of a large fleet, not to retire and watch agents work. Week 3 covers this question explicitly because every cohort asks it.
Mostly group + personal coaching at scheduled cohort cadence, not fully self-paced. The identity work in Weeks 1 and 5 in particular benefits from cohort presence — you're working through the same difficulties as the people around you and reading each other's manifestos and personal-experience posts. Cohort size is capped at 8 per intake to keep coaching intensity high.
Take Week 6 again with the next cohort. Or take the whole bootcamp again — your access doesn't expire. Or trigger the guarantee. The point is the outcome, not the calendar.
Yes — bundled at this price. The cert is also sold separately ($995) for engineers who think they're already cert-ready without the prep. Most people will do both.
Timed, ~3 hours, live observation of the candidate operating 10+ parallel streams while delivering specified outcomes against a provided test repo. Graded on stream count sustained, output quality across streams, coordination hygiene, decision-making about when to bring streams to foreground for QA, and recovery from quota / context / brain-fry incidents. Pass = "Certified AgentHerder" — listable on LinkedIn and recognized across the AgentHerder community. Not-yet = "Apprentice" (includes one free re-take within 6 months). Practice-tested, not knowledge-tested.
Fredrik Wollsén. See "About" below.
Built by Fredrik Wollsén — Agentic Software/Growth Engineer at Augmented Mind. Twenty-two years as a professional software engineer; the last decade making AI genuinely useful, personally and professionally.
Currently running 12–15 parallel Claude Code sessions every day through cctabs — the open-source CLI he built for the job. AI Engineer Principal Tech Lead at F-Secure, where the 4-engineer team he works on has been operating AI-native since August 2025, sustaining 400+ PRs/month against shipping product. Former Claude Code Coach at DNV Cyber. Author of The AI-Native Engineering Playbook: Crawl, Walk, Run, Fly (1,100+ reads). Founder of Remember This and My Transcriber. Previously Mozilla.
On stage at: GOSIM AI Vision Forum Paris, Panel 1 — Agentic AI Systems & Human-AI Symbiosis (May 4, 2026).
Everything in this bootcamp is something the curriculum author does every day with a real fleet against a shipping product. Not theorised. Not from a deck. From actually running the practice.
Six weeks. From running to flying. From 1–3 parallel Claude Code sessions to 10+ sustained. Cert bundled. Outcome-guaranteed.
The Bootcamp opens for its first cohort after the AgentHerder identity launch wraps. Get on the early-access list — we'll email when enrollment opens, and the first 50 enrolments get launch pricing locked.
No spam. One email per milestone. Same waitlist as the certification — you'll be notified about both.
Not yet running parallel agents? Start with the Augmented Mind Bootcamp — the on-ramp that takes engineering teams from walking to running, before you scale into a fleet.