September wasn't a special month. Plenty more got built, but more of what came in was feedback, and that turned out to be the more useful half; it's not that people flooded in, it's that a few people who aren't me started telling me what they ran into, and that's quite a different kind of steering than my own list. Someone also paid for it. Five euros, more a gesture than revenue; someone wanting to give me money more than needing what it buys. It doesn't really matter, and it still does, a little: the zero is gone. The spend next to it is still way too high, and honestly, I'm fine with that.
What Moved
Last month I wrote that I wasn't sure I'd paid back the people testing this. That shifted in September. The people giving feedback now get something for it: a working system, and the improvements their feedback turns into. The feedback is still worth more than I can return, but at least there's a balance now, and that makes it a more interesting trade than asking favours.
The building kept going around it: Node, Next.js and Nuxt apps, Symfony right at the end of the month, a few more Linux distributions you can bring a server in from, and HTTPS on a server without opening ports 80 and 443. But the one I'd put first is that you can now take your server back.
It's quite a big deal to me, even if nobody will use it yet. Once a server is managed by something it's usually close to impossible to get back; the tool is what keeps it up, and pull the tool and things quietly stop working. Here it comes back as a fully configured system, update commands and all; what goes away is us, the management and the looking after. Not being locked in is one of the things I promised, and that only counts if the way out works, so this was close to priority one.
What Didn't
Hetzner took the same node down twice, and the second time, on the 20th, the control plane went with it for 48 minutes. Every customer's app kept serving and recovery happened without me, which is how it's designed; but for nineteen of those minutes I had no way to tell anyone anything, and that bothers me more than the outage.
Monitoring also went blind a few times, and that's its own kind of uneasy; I'd rather know something's broken than wonder. The alerting is being tuned so the important ones reach me without fifteen thousand push messages.
Would I Pay For This Today?
Yes. This website runs on strackt and I don't look at it anymore. I moved it from the preview environment to production with zero downtime, and it just worked. When I change the terms and conditions, a preview pops up that I can send off for review before anything goes live. For me it's pretty much flawless, and that's why I'd pay for it.
Next Month
Keep the pipeline of fixes and improvements running; that part doesn't stop. But October is mostly about the other side, the part that isn't code: outreach, marketing, actually talking about what strackt is and who it's for. Finding people to help beta test, finding people who might be interested, and sharpening how I describe it until the right people recognise themselves. More and more people using it, but only as many as I can actually look after; when something breaks for someone, they should hear back from me, and I'd rather have fewer people who do than more who don't.
One payment down; now I need a lot more of them, and the spend can stay exactly where it is until then.