We map your actual work — the tools, the gaps, the shape your role needs.
Secondary · Open Cohort
Four weeks. Twelve people. You ship something real.
Duration
4 weeks. Twice-weekly live sessions + async between. ~12 seats per cohort.
Who it fits
Working operators (PMs, leads, builders, consultants, founders) who want to learn agent-building hands-on, alongside peers in different industries.
A cohort program for working operators learning to build with agents and modern AI. Mixed-company. Hands-on. By the end you’ve shipped a working project, not just understood concepts.
How it works
Diagnose. Build. Hand you the keys.
One method, three moves: we map your actual work, build your private brain from it, and hand you the keys — owned, portable, queryable from anywhere.
We build your private brain from it — your corpus, indexed and queryable.
It's yours. Reach it from any LLM, at your own subdomain. You own it.
How the cohort works.
Four weeks, twelve operators, a working project at the end.
A mixed-company cohort — a SaaS PM, a healthcare ops lead, a marketing director, a founder, an agency principal. The cross-pollination across industries is half the value.
Twice-weekly live sessions plus async project work. Claude projects, custom GPTs, agent frameworks, MCP — group critique on each project, stuck points worked through together.
By week four you’ve shipped a working agent or AI-leveraged workflow that runs in your actual job — not a toy demo. The recordings and the cohort Slack stay yours.
What the cohort is for.
There’s a gap between understanding AI conceptually — which a podcast can give you — and being able to build with it as a working tool, which requires shipping something. The cohort exists for operators who want to cross that gap quickly, with structure, alongside people who’ll push each other.
The format is hands-on and project-based. The curriculum is calibrated for working operators, not engineers — the math and theory get translated; the focus is the patterns and decisions that matter for shipping.
Mixed-company means you’re learning alongside operators from different industries: a B2B SaaS PM, a healthcare ops lead, a marketing director, a startup founder, an agency principal. The cross-pollination is half the value.
“By the end of week four, you ship a working agent or AI-leveraged workflow that runs in your actual job — not a toy demo.”
Open cohorts run quarterly · ~12 seats
Apply for the next cohort.
The application is a short paragraph on what you'd build. Spots aren't first-come — they're calibrated for diversity of industry and intent. $1,500/seat, 4 weeks, you ship something real.
The four-week curriculum
Twice-weekly live + async project work.
- W1
Foundations
Four building blocks: prompts, projects, agents, brains. Which shape fits which problem. Each participant picks their project.
- W2
Building blocks
Hands-on with the tools. Claude projects, custom GPTs, agent frameworks, MCP. Group critique on each project; we work through stuck points together.
- W3
Architecture + shipping
Decisions that separate prototypes from production. Retrieval shape, error handling, quality calibration, cost. Most ship their first working version.
- W4
Polish + demo day
Final iteration. Documentation. Demo day where each participant presents their shipped project. Recordings preserved as portfolio.
What you walk away with
In the $1,500.
A shipped project
Real, working, useful in your actual job. Not a tutorial follow-along.
The recordings
8 live sessions + demo day. Keep them indefinitely as portfolio + reference.
The cohort
~12 working operators across industries you can reach out to after. Cohort Slack stays open.
The patterns
Working mental model for which tools fit which problems. How to ship AI work as an operator (not as an engineer).
30-day office hours
Weekly 60-min group sessions for the month after for people who want to keep iterating.
Cadence and availability.
Open cohorts run quarterly. Applications open ~6 weeks before each cohort starts, and include a short paragraph on what you’d build — used for both fit and project selection.
The application matters because the cohort works best when everyone is genuinely going to ship. Spots aren’t first-come — they’re calibrated for diversity of industry and intent, which is the precondition for cross-pollination to actually happen.
Common questions
Asked, answered.
Do I need to be technical?
No. The cohort is calibrated for working operators, not engineers. If you can write Excel formulas, you have enough technical background. The tools — Claude projects, agent frameworks, MCP — are configurable without code in most cases. When code is the right move, we use Cursor or Claude Code in ways that don’t require a deep programming background.
What if my project doesn't work?
Most projects pivot — the version you applied with isn’t usually the version you ship. The structure includes explicit checkpoints where you can change direction with peer and instructor input. The shipping criterion is "real and useful," not "matched your original spec."
How much time per week does this take?
4-6 hours/week. Two 90-minute live sessions plus async project work. Some participants invest more in their project; the structure works at 4 hours/week if you stay disciplined.
Can my company sponsor me?
Yes. Most participants are sponsored by their employer. The application doesn’t require it either way; most companies are happy to invest $1,500 in an employee learning to ship AI work.
What if I want to run this internally for my company?
That’s the Closed Cohort— same curriculum, run privately for one company, customized to the company’s use cases.
Adjacent paths
Where this connects.
Start with the Diagnostic Call.
The Diagnostic Call commits to nothing. If this install isn’t the right shape for your work, we’ll tell you what is.
As featured in
