Learn agentic workflows.
Ship production code with AI agents.
A hands-on course where you learn to plan, execute, verify, and correct work alongside AI agents — in real codebases with real constraints. Tool-agnostic. Not a prompt pack. Not a tool tutorial.
What you'll learn
This course teaches a repeatable agentic workflow: spec → plan → execute → verify → correct. You'll practice it on real tasks in real repos — tickets, PRs, refactors, and bug hunts.
By the end, you'll have a personal playbook and reusable templates (specs, task briefs, PR checklists) you can apply immediately at work.
How to structure multi-step AI work: planning and execution loops, review and verification passes, search strategies for large solution spaces, and when to split work across chat, IDE agent, and CLI.
Turn vague requirements into bounded specs with acceptance criteria an agent can actually follow.
Keep agents effective with context budgets, canonical reference files, and reset protocols.
Tests, lint, format, diff review — build evidence-based checkpoints so nothing ships broken.
Run the full workflow on your own repo: ship a real change with verification artifacts and a postmortem.
How the cohort works
A 6-week live cohort for mid-to-senior engineers already shipping production code. Each week has a module, a hands-on exercise in your own repo, and a group review session.
Weeks 1–2: mental models, failure modes, work decomposition, and spec writing.
Weeks 3–4: agentic workflow patterns, context management, and multi-step execution.
Week 5: guardrails, evidence-based review, and building verification into your workflow.
Week 6: apply the full workflow to your repo — ship, verify, and document everything.
Mid-to-senior engineers, tech leads, and staff+ already working with CI, PRs, and tests.
A real repo you're actively working in. The messier the better — that's the point.
Works with Cursor, Windsurf, Copilot, Claude, or whatever you're using.
Curriculum
6 weeks focused on workflow systems — not tools, not prompts. Every module ends with an artifact you can reuse.
Apply the full workflow to a real repo. Ship a change with tests and verification artifacts, then write a short postmortem.
- Spec + acceptance criteria
- Plan + task briefs
- PR checklist + verification evidence
- Shipped change + postmortem
Use AI as a junior dev, pair, reviewer, and refactor assistant — without letting it steer architecture.
Turn "build X" into bounded work: constraints, definition of done, and a plan an agent can execute.
Multi-step workflows, review loops, and search strategies — choose patterns by task shape and risk level.
Keep agents on track with context budgets, canonical reference files, and reset protocols.
Tests, lint, format, diff review protocols — nothing ships without evidence.
Run the full workflow on a real repo: ship, verify, and document the system.
Recommended reading: security, secrets, and sandboxing — containment is part of correctness.
FAQ
Short answers to the questions engineers ask before enrolling.
Is this an ML course?
No. It's engineering workflow: how to run AI in real repos with verification, guardrails, and review discipline.
Is this a tool tutorial (Cursor/Windsurf/Copilot)?
No. We're tool-agnostic. We teach decision-making and workflows that transfer across tools and stacks.
Is this for beginners?
Not for beginners. This is for engineers already shipping production code with tests, PRs, and CI.
Do I need a real repo?
Yes. The course is designed around real constraints: legacy modules, flaky tests, CI friction, and scope creep.
What if my repo is messy or tests are flaky?
That's normal. We treat it as a constraint: tighten scope, add guardrails, and require evidence before shipping.
What do I actually produce?
A workflow playbook + reusable checklists/templates, plus a capstone change with verification artifacts and a short postmortem.
Join the first cohort
Spots are limited. Reserve yours in 2 minutes — submit your enrollment from your dashboard after signup.