Cohort 01 · Spring 2026 · Now Open
INDEX / 02 · THE ENGINEERING TRACK

The GenAI
Builders.

The Agentic Engineering program — from vibe coding to production-grade multi-agent systems. Ship 9 real projects. Earn a certificate that means something.

Reviewed 4.9 / 5 · 300+ builders
★ ★ ★ ★ ★
Next cohort: May 15
Duration
6–10 Weeks
Format
Self-paced + Discord
Output
9 Shipped Projects
Certificate
Verifiable
Start Building →See the 9 modules
Scroll
↓
LangGraph ✦ CrewAI ✦ Cursor ✦ Lovable ✦ Bolt.new ✦ v0.dev ✦ Replit Agent ✦ Supabase ✦ Vercel ✦ n8n ✦ MCP ✦ Claude Code ✦
01 · Premise

Basic GenAI is table stakes. The real leap is Agentic Engineering — orchestrating autonomous agents that plan, reason, use tools, self-correct, and ship work with minimal supervision.

01.1
Architect, not typist.
Karpathy delegates 80% of his coding. You will too. Your job is to design, supervise, and verify — not to hand-write for loops.
01.2
Orchestrate, don't prompt.
One prompt is a toy. A team of agents planning, acting, and checking each other is a product. We teach the second thing.
01.3
Evaluate or hallucinate.
The difference between a toy and production is measurement. Evals, tracing, and guardrails: the boring stuff that makes you employable.
01.4
Ship, not vibe.
Vibe coding is step one. Step two is shipping: deployments, observability, and systems that survive real users.

From vibe coding
to production.

A path from “I can build demos” to “I can ship systems”.

01 · Foundations

Learn the loop.

Prompt + context engineering, tool design, and the iteration loop that makes agents reliable.

02 · Systems

Build the stack.

RAG, vector DBs, multi-agent graphs, MCP integrations — the real primitives of agentic engineering.

03 · Production

Ship the product.

Deploy, eval, monitor, and defend your work in interviews. A portfolio that proves you can ship.

+ This is for

If you want to ship what teams are hiring for — now.

  • 01Developers leveling into agentic engineering
  • 02Technical founders shipping MVPs without a team
  • 03PMs who want to prototype, not just spec
  • 04Engineers chasing the #1 AI skill of 2026
  • 05Beginners willing to go deep, not skim
− This is not

Sit this one out if…

  • 01You want a generator, not an engineering discipline
  • 02Debugging makes you want to quit
  • 03You're allergic to reading docs or APIs

Nine modules.
One discipline.

Every module ships a portfolio-ready project. Click any row to expand — what you learn, what you build, and how it compounds toward the capstone.

What you learn
  • The delegation mindset — Karpathy’s 80% rule
  • Vibe coding vs. hand-coding — when to do which
  • The iteration loop: spec → generate → evaluate
  • Reading AI-written code critically
  • Your new daily stack
Module 1
You shipA working landing page shipped entirely by vibe coding.
What you learn
  • Anatomy of a production prompt
  • Context windows — what to include, what to cut
  • Few-shot, chain-of-thought, and self-critique
  • Prompt evaluation and A/B testing
  • Versioning prompts like code
Module 2
You shipA prompt library + eval harness for your domain.
What you learn
  • The three major APIs and when to pick which
  • Structured outputs (JSON mode, schemas)
  • Function / tool calling fundamentals
  • Streaming, retries, and error handling
  • Token economics and cost guardrails
Module 3
You shipA multi-model CLI that routes queries by task type.
What you learn
  • Embeddings intuition and cosine similarity
  • Chunking strategies that don’t ruin recall
  • Pinecone, Weaviate, pgvector, Supabase
  • Hybrid search and reranking
  • RAG evaluation — the metrics that matter
Module 4
You shipA production RAG system over your own docs.
What you learn
  • The ReAct loop: reason → act → observe
  • Tool abstraction and the tool belt pattern
  • Memory: short-term, long-term, episodic
  • Self-correction and retry strategies
  • When agents fail — and how to debug them
Module 5
You shipA single-agent research assistant with web + file tools.
What you learn
  • LangGraph — stateful agent graphs
  • CrewAI — role-based collaboration
  • Supervisor / worker / critic patterns
  • Handoffs, shared memory, and human-in-the-loop
  • Observability with LangSmith
Module 6
You shipA multi-agent pipeline that plans, executes, and reviews itself.
What you learn
  • What MCP is and why it matters
  • Building an MCP server
  • Connecting Claude Code, Cursor, and custom agents
  • Secure tool exposure and auth
  • Real-world MCP patterns (files, DBs, APIs)
Module 7
You shipA custom MCP server that connects your tools to any agent.
What you learn
  • Deploying to Vercel, Supabase, Cloudflare Workers
  • Evals: offline, online, and LLM-as-judge
  • Tracing with LangSmith / LangFuse
  • Cost, latency, and caching strategies
  • Security and prompt-injection defense
Module 8
You shipA deployed agent with dashboards and alerting.
What you learn
  • Scoping an agentic product
  • Architecture: graphs, memory, tools, guardrails
  • Evaluation harnesses you can defend
  • Deployment and observability
  • Presenting to technical stakeholders
Capstone
You shipA live, monitored, multi-agent product + a case study video.

The stack.

Not a tool list — an opinionated engineering stack. We teach you which, when, and why. Frameworks change; architectural intuition doesn't.

01IDEs & Vibe Coding
CursorDaily driver
Claude CodeAgentic CLI
LovableRapid MVP
Bolt.newPrototype
v0.devUI scaffolds
02Models & APIs
AnthropicReasoning
OpenAIGeneral
GeminiLong context
GroqLow latency
Together AIOSS models
03Agents & RAG
LangGraphAgent graphs
CrewAIRole-based
PineconeVector DB
pgvectorPostgres
MCPTool protocol
04Ship & Observe
VercelDeploy
SupabaseDB + auth
LangSmithTracing
n8nAutomation
Replit AgentHosted agents

You leave
with proof.

Not a certificate. A working stack, a filled portfolio, and a new identity on LinkedIn that actually means something.

01

A production skillset.

Prompt engineering, RAG, agents, multi-agent graphs, MCP, deployment. The full agentic stack — not toy demos.

02

A portfolio of 9 shipped projects.

Every module is a deployable artifact. GitHub-ready, demo-ready, and in a stack engineering managers recognize.

03

A live multi-agent capstone.

A real agentic product, deployed, monitored, evaluated — and defendable in a technical interview.

04

10× velocity, not 10× hours.

Workflows, templates, and agents that do the boring 80% — so you design, review, and ship at a different tempo entirely.

05

Frameworks that compound.

Eval harnesses, agent templates, MCP servers, tracing dashboards — reusable infrastructure for every project that follows.

06

A new title.

Agentic Engineer. The #1 AI skill of 2026 — and the role that didn't exist three years ago.

Cohort 01 · Closing Soon

Engineer the
machine.
Outship the rest.

6–10 weeks. 9 shipped projects. One real capstone. Seats are capped so reviews stay hands-on.

Enroll in Builders →See Creators Track
Starts May 15, 2026
Self-paced · Kept Forever