AI Agents
Four AI agents, one founder
As well as my day job, I’m also a solo founder running a music tech startup with four AI employees.
Not hypothetically. Right now, today, this is how SquareWave Studio operates.
What happened while I was making dinner
A beta tester emailed support about some bugs in EP-PatchStudio (my desktop app for Teenage Engineering devices). Within minutes, Patch, my operations agent built on the Claude Agent SDK, checked the inbox, read the emails, classified them, ignored unimportant ones and created detailed GitHub issues for the bug reports.
GitHub Actions picked up the new issues and spun up EP-PatchStudio Dev (Claude Code) to investigate the problems. It found the root cause for most of these, wrote fixes and opened PRs. For the ones it was not confident with, it saved a summary of what it thought the issue was in a comment on the bug.
I then got a notification on my phone with a summary.
Beta tester emails to pull requests. No human in the loop.
The team
-
Patch: operations and analytics. Monitors the inbox, tracks metrics across PostHog, LemonSqueezy and Brevo, monitors social media, forums and Discord, triages support into GitHub issues, drafts emails and social media posts, runs daily briefings.
-
EP-PatchStudio Dev: investigates issues, writes fixes, works on feature enhancements and raises PRs for the Tauri desktop app.
-
SquareWave WebDev: maintains the marketing sites and download infrastructure on Cloudflare.
-
SquareWave AgentDev: builds and maintains Patch itself. Yes, one agent develops the other.
The whole setup runs on Claude Agent SDK and Claude Code. Patch checks the inbox every 2 hours, collects business metrics every 6 hours and sends me a daily briefing. Most mornings I wake up to a summary of what happened overnight and a list of PRs to review.
What I still do
I’m not sharing this to say AI is replacing developers. I’m still the one making product decisions, talking to customers and designing features. But the operational overhead of running a software company (the ticket triage, the metric tracking, the routine bug fixes) is handled.
I’m still pre-launch and in beta testing, but that’s exactly why I built this team. When it’s just you, everything needs to be as automatic as possible.
What this means for how companies operate
The operational leverage available to small teams right now is unprecedented. The tasks that used to justify hiring your first support person, your first DevOps hire, your first marketing coordinator can now be handled by agents running on structured workflows with human approval gates where they matter.
This isn’t about replacing those roles. It’s about not needing to fill them before you’ve validated whether the product has legs. The cost of testing an idea just dropped by an order of magnitude. For solo founders and small teams, that changes the calculus on what’s worth attempting.