I Went to Bed While Claude Was Still Working. I Woke Up to a Fixed App.

My eyes swam at the 5 page essay of bugs.

The mouse cursor jolted to a new line when the overlay was open. The ember wasn’t blooming on the third pomodoro session.

I was watching Resident Alien with my girlfriend and throwing small prompts at Claude. It kept telling me I needed to eyeball the changes.

Totally get it. Totally didn’t want to do it.

I remember asking Claude Code, “Is there a way you can do all of this for me?”

Thinking…

It pasted a wall of text. But the first line was all I needed to see.

“Yes.”

I didn’t even watch it start.

I Watched It for a Second and Looked Back at Harry.

It was getting later and I’d queued up another episode of Resident Alien.

Harry was talking to Max and called him a “shit pocket.” I relaxed in my chair and laughed. And while this was happening Claude was controlling my computer and stress testing my app.

At some point I glanced at my screen. The edges were pulsing. My mouse jolted up on its own, opened Lannair, and started clicking and typing. I watched it for a second and looked back at Harry.

If this didn’t work, I’d have to manually test my app. Not a big deal. It would be annoying and time consuming.

But if it worked I’d be able to work on my app while I slept. I wanted to go to sleep as soon as possible so I could wake up and see what happened.

So I went to bed.

It Fixed What I Asked. Then It Found More.

It was 4 am when I woke up and stumbled out of my bed.

I sat on a pile of clothes I’d left on my chair from the night before. Had it worked? Or had the tests failed in the middle of the night?

I opened up Claude Code and was greeted with an even longer essay than last time. But this time it was about bug fixes.

Claude documented the cursor and pomodoro problems. Went through edge cases for the markdown. And fixed the issue with typewriter mode’s cursor jumping up a line when the overlay was opened.

But as I scrolled through the fixes, I found new bugs it had spotted on its own. Bugs it couldn’t have found without running the app live.

Claude added them to a new list and kept going. One was a memory leak when closing out the app.

Claude fixed more than 30 issues in 2 hours. I would have been at my desk until noon.

Twenty Seconds. Hours of Work.

I didn’t know how to set this up. So I asked Claude.

It told me to use Cowork. I didn’t overthink it. I just said: “Based on the stress test you’re recommending, write me a prompt that will let Cowork handle everything. And if you find any bugs, update them along the way.”

That was it. Twenty seconds of typing.

What came back was 30 fixes. Including bugs I hadn’t put on the list.

Most people hand off something like “Claude, fix the issues you find” and wonder why the results are vague. Vague brief, vague output. That’s not a handoff. That’s an abdication.

The difference is specificity and room to move. Tell Claude exactly what you want done. Then give it permission to make calls along the way.

You put in the twenty seconds. Claude handles the hours.

You’re the designer. Claude is the developer.

The thing about building Lannair alone is that I want to give it everything. I lose track of the hours. I forget to stop.

The problem is I’m still one person. Limited on time. Limited on funds. I can’t hire anyone to share the load.

But I woke up one morning to 30 fixes I didn’t write, in hours I was asleep.

When your hours stop being the ceiling, everything changes.