AI Enhanced Dev

Build (web) apps with AI agents — without the disadvantages.

Length Two weeks
Commitment ~1 hour a day
Outcome Build and deploy, slop-free

A two-week program where I share how I actually build applications with AI — and how I keep the slop out.

We're all building with AI now. Some of us are not writing or reading any code anymore; some of us are a bit deeper in the code. And we're all fighting AI's tendencies to add slop, get itself lost in dead ends, or add endless amounts of complexity.

In this 2-week program, I (Max) share everything I've learned over the last 2 years of building (web) applications with AI.

The thing is: Over those last years, how you efficiently work with AI kept on changing — from auto-completion to chat to bad agents to much better agents (who can still spit out tons of slop!). And it keeps on changing.

In this program, I share my current approach from A to Z, and also how I quickly adapt to new models, patterns, and tools.

So this is not a prompt collection. Also not a course about some library or framework. It's instead an in-depth program: what you research before you plan, what you plan before the agent writes a line, which rules and gates the project enforces on its own, how to pick a proper tech stack, how to orchestrate one or more agents, and how you review what comes back. All applied to a real project which we build as part of the program.

Let AI accelerate you — but skip the slop!

Using AI does not mean that you need to drop or lower your standards. None of this is about accepting worse code because it arrived faster. It's about leveraging your skills!

What we cover

Two weeks and one real project — starting with the research that happens before any code, ending with an application that is deployed.

Research and planning, first What you find out before you plan, and what you plan before the agent writes a line. No waterfall and no giant upfront spec — plans that stay alive as the project moves.
A project set up to hold a standard The skeleton, the linting, the formatting and the anti-slop rules, plus the best practices you already have — applied where an agent would otherwise cut corners.
Rules that keep an agent on track Writing and using proper agent skills, enforcing tests that are truly helpful, and how to steer the agent to stay on track.
Working with agents — one, then several Prompting and planning (without plan mode), judging what comes back without reading every line, then worktrees and running more than one agent without them tripping over each other.
A demo application, end to end We build one real project together with SvelteKit, Effect, Alchemy and Cloudflare — from an empty repository to something deployed and running.
Debugging, optimising, maintaining Using AI on a codebase that already exists, optimising in a loop, and keeping the rules, skills and docs from rotting once the project is real.

Follow along however you work. In the program I work on a remote machine through a terminal UI, and I show you exactly how that is set up. Nothing here depends on it — everything works the same on your own machine. And of course the learnings apply to other tech stacks just the same!

Why I'm running this

The problems we face

We developers can all feel it: Everything's changing. AI seems powerful, frightening, and weirdly limited at times. On social media, everybody seems to be pulling ahead.

I've certainly been there! But I've also found strategies and approaches that let me truly be more effective with AI. That allow me to build faster and more.

And in this program, I'm sharing them with you. Step by step, from developer to developer.

What slop-free means

Code that does not just work but forms a solid foundation. Aligned with best practices, typed, tested, reviewed, and shaped by clear rules.

Code that's following clear guidelines and rules, and that doesn't grow into a complexity monster.

Who this program is for

Every developer who feels they should be getting more out of AI than they do right now — on your own or in a team, on a new project or on a codebase that has been around for years. The stack we use here does not matter; what you take away transfers to whatever you build with. You do need to know how to build (web) applications, but you do not need to be an expert at it.

Want a practical starting point? Read How to actually get the most out of Claude Code for the prompts, context, skills and models I'd focus on first.

What you get

Preparation resources Short mini courses to work through before we start: enough Effect, Alchemy and Cloudflare to understand the code the agent writes, and the platform it deploys to. Day one is not day zero.
The program curriculum Ten days of material, unlocked on the start date. You work through it at your own pace.
Twice-weekly async Q&A You send your questions upfront and I answer them twice a week, as video and/or text. There are no live sessions to attend.
Lifetime access Everything stays yours after the two weeks are over.

Two weeks, about one hour per day Tentative

The mini courses before we start, ten days of material, and twice-weekly check-ins throughout. The two weeks unlock on the start date; you go at your own pace.

Pre-learning — three mini courses

Available before we start. Short on purpose: enough of each to understand the library, the platform, and the code the agent writes — not a full course in any of them.

Effect: Understanding the library and the code the agent writes: generators, typed errors, services and Schema. No category theory.
Alchemy: Understanding the infrastructure the agent writes: stacks, resources, typed bindings and stages.
Cloudflare: The key services and features we use, and how Alchemy works with them.

Week 1 — Research, planning, first slice

What you research, how you plan, and the setup that holds the project to a standard on its own — then a first slice of the app, actually running.

Introduction and the plan: Where building with AI actually fails, why a spec written up front drifts, and working out what you are building: the key features, the parts that are genuinely hard, the roadmap and the rough overall plan.
The stack and the agent setup: Hosting and constraints, the tech stack, a testing strategy, the interfaces and schemas everything else hangs off — plus one source of truth for the project, agent skills, AGENTS.md, enforcing agent behavior, and choosing an agent harness.
Project setup and first slice: The skeleton, the linting and anti-slop rules, ephemeral deployed test environments, and a first thin slice running end to end.
Working with agents, one task at a time: Creating and implementing plans with agents - without micromanagement, but also without ugly surprises.
Code reviews: Judging a change without reading every line: automated first-pass gates, risk-based triage, keeping changes small, and working in layers.

Week 2 — Building, shipping, improving

Working through the rest of the app with one or more agents, then cleaning up, deploying, and keeping a real project improving.

Building the app, with worktrees: Working through the bulk of the app — the grunt work — with worktrees, one slice after another, and letting an agent measure its own work so you stay out of the complexity trap.
Building with more than one agent: Building the rest of the app with more than one agent: how many a task actually supports, handoffs that survive a context boundary, and what a fleet really costs you.
When the agent goes off the rails: Cleaning up AI-written code, and stopping an agent that's about to go off the rails — too much complexity, the wrong direction, or code you would not keep.
Deployment and observability: Getting the app deployed and seeing what it actually does once it's out — with help of AI.
Ongoing improvement: Improving a project that already exists — in a loop with AI, and manually. Keeping the rules, skills and docs from rotting.

Twice-weekly check-ins

Questions in as they come up, answers out twice a week. No live sessions to attend.

Your questions: Send them while you work through a day or a mini course.
My answers: A combined video and/or text, twice a week, covering what the cohort asked. Yours to keep.

My promise

After completing this program, you're ready to build your own slop-free applications with help of AI. No matter the exact tech stack.

Or you'll get your money back!

Meet your instructor

Maximilian Schwarzmüller Founder of Academind

I'm Max. I started coding at 13 and I have been at it for over two decades. Somewhere along the way I started teaching it too — over three million students, on Udemy and on my own platform. I also wrote a book about React.

For the last two years almost everything I build has been built with AI in the loop. This program is what came out of that: not theory, but the way I work right now - based on hundreds of hours of learning and building.

One price

$449

The full two-week program
All preparation resources
Twice-weekly async Q&A
Lifetime access to everything in it

I am planning a limited number of PPP spots (prices on request) and team pricing for three or more licenses. More details once enrollment opens up.

Questions you may have

Do I need to know Effect, Alchemy or Cloudflare already?

No. There will be mini courses on each before we start. I do assume solid TypeScript and web development experience.

Which AI tools do I need?

Claude Code, Codex, or whatever your favorite AI agent is, with a paid plan for at least one. Other agents work too — those two are simply what the examples use. In the program I work on a remote machine through a terminal UI and show you how it is set up, but you can follow along entirely on your own machine.

Are there live sessions?

No. You send your questions upfront and I answer them twice a week, as video or text.

How much time should I plan?

About an hour a day for two weeks. Roughly ten hours in total, at your own pace.

Is the curriculum final yet?

No, the curriculum is tentative but the general outline and what you'll learn is final. You will learn how to build slop-free applications with AI, using the tech stack mentioned on this page.

When does it start?

I have not set the exact dates yet, but expect November/December. Leave your email and you will hear first! I can't guarantee it'll be before GTA 6.

Can I enroll already?

No, at this point, you can leave your email to get notified once enrollment opens up. No obligation to pay or enroll!

What does it cost?

$449, and that is the whole program. PPP spots and team pricing will be announced once enrollment opens up.

Do you offer refunds?

At this point, enrollment is not open yet, but once it is, I will, yes. I'm 100% convinced that this program is awesome, so I'm more than happy to offer refunds if you don't like it.

Can my company pay?

Yes. I invoice with a VAT id, and I quote personally for three or more licenses.

Do I keep access after the two weeks?

Yes. Lifetime access to every program resource.