Three months from idea to shipping software

Three months ago I had an idea. Today I’m shipping software.

In mid-November I bought a Teenage Engineering EP-40 and immediately ran into the same frustration every EP device owner knows. The official sample management tools are slow, online-only and painful. I’d already built a free web tool for their OP-XY device that grew to 3,500 monthly users, so I knew there was demand. But this time I wanted to build something bigger.

The problem? Teenage Engineering don’t publish their device communication protocol. No documentation. No SDK. No API.

So I reverse-engineered it.

I used the browser console to log every MIDI SysEx message the devices sent and received. Then I fed those raw hex dumps to AI and together we decoded the entire protocol into working documentation. Every command, every data structure, every edge case. All done legally under the EU Software Directive’s interoperability exception.

From there I built EP-PatchStudio. A full native desktop app for managing samples, creating multisamples, auto-sampling hardware synths via MIDI and backing up devices. Cross-platform: Mac, Windows, Linux. Built with Tauri and Rust.

What made it possible

Here’s the thing though. I have a full-time job. My partner and I run an animal sanctuary where we care for 60+ rescue animals. There is no world where I could have built and shipped a commercial software product in three months without AI tools. Not a chance.

AI didn’t just help me write code faster. It helped me learn entirely new disciplines. I taught myself sales funnel design, email campaign strategy, conversion optimisation, SEO, freemium pricing models and product launch playbooks. Things I’d never done before. I built a payment system with LemonSqueezy, analytics with PostHog, email campaigns with Brevo and a full marketing website.

Then I went further

I built an AI agent called Patch using the Claude Agent SDK to actually run the business. It monitors support emails, triages bug reports into GitHub issues, tracks analytics across every platform, scans Reddit and forums for user feedback and even drafts marketing content trained on my brand voice. I now have four AI agents operating as my team: Patch for ops, a dev agent that investigates and fixes bugs automatically, a web dev agent and an agent dev that builds and maintains the others.

One of my agents literally develops another one of my agents.

Three months. From an idea to a company called SquareWave Studio with a product, a brand, a sales infrastructure, an AI-powered operations team and real users.

What this actually means

If you’re sitting on a side project thinking you don’t have the time, the skills or the resources to ship it, the equation has changed. The tools exist now.

The gap between “I think this should exist” and “here it is” has collapsed. Not because AI writes perfect code. But because AI compresses the learning curve on everything surrounding the code: the business model, the marketing, the operations, the infrastructure. The parts that used to require either deep experience or a team.

What used to be a reason to not start is now just a thing you solve on a Tuesday evening.