I already run a software factory. It lives on a VPS, a stack of Claude Code and ChatGPT subscriptions feeds it, and on a normal day it opens hundreds of pull requests across my projects. Nothing about it is broken.
So why spend a hundred days on Devin? Partly because I wanted to try something new, and partly because of a question I can't answer from the outside: can I hold that pace when another company runs the orchestration and I'm working with a smaller quota than the one I've built up? Moving everything to Devin would mean giving up most of the headroom I have today. What I'd get back is a managed layer I don't have to maintain myself, and that trade deserves a real test instead of a guess.
I'm keeping a journal for the full 100 days, and what follows is day one.
Why the bots tipped it
Devin Bots are what made me go all in. I could rebuild my fleet as bots, each one pointed at a workflow or a product I'm already running, and invite them into Slack where I work anyway.
"All in" comes with an asterisk, though. Claude Code, GPT and a few other models keep running in the background, because I'm fairly sure that without them I'd burn through the max Devin plan's quota in half a day.
Day one was setup
I installed the whole ecosystem: the CLI, the desktop app, the Slack application and the first bots. A few of my workflows moved across, but most of the day went into the interface itself, which I read option by option and session by session, opening every sidebar to see what was in there.
It's a clear interface. I'm just not fluent in it yet, and the intuition isn't there.
Where it hurt was session tracking. With twenty or thirty sessions running at once, I couldn't get one view of all of them and their history, so the same questions kept coming back. Which one produced a pull request? Which one needs me? I clicked through sessions to find out, and that doesn't scale to a fleet. Slack may fix it once the bots report there; I'll know in a week or two.
By the evening I had around fifteen bots invited to Slack, shaped after the system I already use. I haven't switched the old one off, so for now I'm running a hybrid.
The first thing that's different
Devin shows me QA scenarios as video.
I work on several projects at once, and walking through a full scenario by hand in each of them is time I mostly don't have. Most tools hand me screenshots, and a screenshot can't tell me whether an animation stutters or how the flow moves between steps. A recording can, so I check the whole scenario properly without clicking through it myself. Of everything I touched on day one, it's the feature that sets Devin apart from the other tools I use.
18% for setup

The quota scares me. Setting everything up cost 18% of the weekly allowance, and that's with 80 or 90% of the day's real work running on my other subscriptions. Devin barely carried any load and still took nearly a fifth of the week.
The reset is three days away, so for a setup day I think that's fair enough. I'll probably spend the full 100% before then, and that's when I find out what a week of real work costs.
One small minus
There's a promotion going through Poke that gets people free accounts. I'd paid one day earlier, which is just bad luck.
I tried it anyway. If you're not an Apple person, Poke steers you to WhatsApp, so I registered everything there on my phone, and only after all of that did it tell me I couldn't get the promotion. That's a bit of clickbait and it wasn't fun. I'd probably have signed up for Devin regardless, so all it cost me was the time on the phone.
What's next
I think I have two more days of setup before the system works the way I need it to. I'm running on my own cloud and on Devin's, and trying not to drain the weekly quota while I do it.
First impression: positive. Ask me again when the quota hits zero.



