I built an AI that tells builders exactly what to do next. It's in beta. Real users are already paying for it.
Next.js · TypeScript · Supabase · Anthropic API · Prisma · Stripe · Vercel AI SDK
The Insight
With AI tools like Cursor and Claude, anyone can now build software. The bottleneck is no longer technical — it's deciding what to build, and actually executing it.
I felt this personally. Handling my internship, school projects, and personal projects simultaneously, I'd come home after work with limited energy and spend most of it figuring out what to work on instead of just working. Not because I lacked motivation — but because managing multiple projects at once creates a specific kind of paralysis: everything feels urgent, nothing feels clear.
I talked to other entrepreneurs I knew. Same problem, every time. Everyone was juggling too much and losing time to indecision. And when they turned to AI for help, the setup cost was too high — giving Claude or ChatGPT the full context of your projects, your goals, your constraints takes time, and the moment you switch conversations, you're starting over. On top of that, it's just text. There's no way to check off a task, track progress, or let the AI know where you actually stand without explaining it manually every time.
So we reckoned that the solution had to be an AI that already knows you. Not a tool you consult — a coach that follows you. One that holds the full context of everything you're building, knows where you stand today, and tells you exactly what to do next.
The Product
Harvey is an AI coach that follows you through everything you're building. It knows your projects, your constraints, your schedule, and where you stand — so every time you open it, the context is already there.
Managing multiple projects at once, coming home after a long day with limited energy, starting something new from scratch — Harvey handles all of it. It tells you what to work on tonight based on your goals and your current progress. It follows you through a new project from day one. It adapts when life gets in the way.
What makes it a real coach is that it learns from you. If you consistently skip tasks on Mondays, it stops scheduling them. If your coding sessions always run longer than estimated, it adjusts. If you're most productive on Tuesday mornings, it protects that time. The longer you use it, the more accurate it gets.
Getting started takes five minutes. You describe your projects, your constraints, your available time — and Harvey is ready. From that point, it follows you. Daily check-ins, task-level coaching, behavioral patterns building over time. The more you use it, the less you have to think about what to do next — and the more you just build.
Smart onboarding
Describe your project in a conversation. Harvey extracts your goals, constraints, and schedule in real-time and builds your first week automatically.
Per-task coaching
Every task has its own dedicated Harvey thread. Full project context pre-loaded. No re-explaining, ever.
Daily check-in
Every time you open Harvey, it knows where you stand — yesterday's completions, today's priorities, your patterns. It tells you exactly what to do next.
How I Built It
I built Harvey because I was the first user. Managing my internship, school projects, and personal projects simultaneously, I was already using Claude as a makeshift project manager — typing long context-heavy messages every evening to figure out what to work on. It worked, but it was painful. Harvey is what I wanted that to actually be.
Before writing a single line of code, I talked to entrepreneurs I knew. Every conversation confirmed the same thing: the problem wasn't unique to me. Everyone was losing time to indecision, juggling too many things at once, and struggling to keep AI tools up to date with where they actually stood. I knew I was building something people needed.
I applied the main lesson from Flyte — start talking about what you're building before it's ready. I started raising hype early, sharing the vision with the people I'd spoken to. By the time I launched the first version — barely polished, but enough to solve the problem — they were already waiting. They became paying users immediately. €20 for three months of beta access. The product worked from day one.
The full product — auth, onboarding, schedule generation, task tracking, per-task chat, Stripe, deployed on Vercel — built solo, in evenings and weekends, while working full-time.
What's Next
Harvey is currently in beta with a small group of paying users — entrepreneurs who came looking before the product was even finished. I use it myself every day. Every week I'm iterating based on real feedback from real builders.
The next layer is deeper pattern recognition — Harvey noticing how you actually work over time and proactively adapting to you, not just tracking what you tell it. The goal is a tool that gets meaningfully smarter the longer you use it.
The longer vision is bigger: not just helping builders execute, but helping them decide what to build at all. I think the hardest problem for builders isn't execution — it's knowing which direction to execute in. That's the problem I want Harvey to eventually solve.
Signals
Inbound demand
Beta users came before the product was finished. Entrepreneurs who recognised the problem reached out directly — not cold signups, but people who were already looking for exactly this.
Paying from day one
The first version was barely polished. It was enough. Every beta user paid immediately — €20 for three months of access. The problem was real enough that people didn't wait for a finished product.
Solo, in production
Auth, onboarding, AI coaching, task tracking, per-task chat, Stripe payments — built alone, evenings and weekends, while working full-time. Deployed and running in production.