Why I Still Ship as a CTO

Should a CTO still write code?

Yes. Not all day. But yes.

I've led engineering teams since 2002. Small ones, and a technology team of about 500. The bigger the org, the easier it is to drift from the work. You start running things off status updates and slides, and your sense of what's hard goes stale.

Building keeps me close to it. When I ship something myself, I know what a week of work actually looks like. I know which estimates are soft. I know what the tools can do today, not just what a sales deck says.

That matters more now. AI tools change fast. I can't make good calls about them for a whole company if I've only read about them. So I use them to build real things, and I find out where they break.

It also earns trust. Engineers can tell when the person setting direction has never touched the system. I'd rather argue about a design with the code open.

What have I actually shipped?

Two products and a few open-source packages.

Agent Console

A desktop app for running a team of AI agents. Each agent gets a name, a role, a working folder, its own terminal and a daily budget. They share a task board, a memory and a handoff system.

I started it on nights and weekends in June 2026. The record:

• 352 commits. All mine.
• First commit June 20, 2026. Latest October 29, 2026.
• About 35,700 lines of TypeScript.
• Version 1.4.0.

I use it every day. By my own estimate it has roughly 4x'd my output. About a dozen people across Welby use it daily too. More on the Agent Console page.

Markus

Welby's clinical AI teammate for nurses, live at markus.getwelby.com. I designed and built Markus and wrote most of the code. A few Welby engineers contributed to the backend. The commit history backs that up: all 171 frontend commits and 242 of the 303 backend commits are mine.

A nurse asks a question in plain language. Markus pulls the real patient record, shows each step it takes, and checks every clinical value against the data it actually fetched. If something is missing, it says so instead of guessing. In clinical work, a confident wrong answer is worse than no answer. More on the Markus page.

Open-source packages

I maintain a few packages on npm, including @reyco1/nestjs-stripe, @reyco1/nestjs-firebase, @reyco1/create-angular-tailwind, @reyco1/release-notes and @reyco1/bump. Small tools I needed, so I published them.

How does a CTO find time to build?

Some of it is nights and weekends. That's how Agent Console started.

The rest is leverage. I built Agent Console so agents could take work off my plate, and now they do. Research, drafts, first passes at code. I review, decide and ship. That's where the 4x comes from.

A few habits keep it sane. I build things I'll use myself, so I hit the bugs before anyone else does. I ship small, because a version that works this week beats a perfect one next quarter. And I keep a record: commits, a changelog, release notes. That's why I can point to 352 commits instead of just saying I code.

When shouldn't a CTO write code?

It goes wrong in a few predictable ways.

• You're the bottleneck. If a release is waiting on your pull request, hand it off.
• It's an escape. Coding is fun. Hiring, budgets and hard conversations are less fun. They're still the job.
• You skip the process. Your code gets the same review and tests as everyone else's. The title doesn't make it safe.
• The team stops owning things. If people wait for you to fix it, you took something from them.
• You're doing it to prove you still can. That's ego, and the team will see it.

If any of those are true, step back. Leading the org comes first.

What does a CTO who still ships look like?

Short version: run the org, and keep your hands in the real work.

For me that's leading technology at Welby during the day and still writing the code for products I build. Markus and Agent Console are what that looks like in practice. The Agent Console code is private, but I'm happy to walk through it live. Email me at hello@reyco.me.