October was supposed to be the month I got out more: talking, writing, putting strackt in front of people instead of in front of my editor. I did some of that, just not nearly as much as I wanted; the product kept pulling me back, and most of the time it had a pretty good reason to. The product got better. I didn't get out enough.
What Moved
The one I'd put first happened on the 21st. PHP shipped a fix for a remote code execution bug, and by the afternoon every server on strackt was running the patched version; I wrote about that day separately, and the short version is that the fast part was the preparation. The day after, release channels went live: a change like that now reaches a few canary servers first, then early, then everyone, and I can halt it halfway if something looks off.
Around it, loads of work that I'd call making things true. ARM servers work end to end now, and a running environment can move between x86 and ARM. A server that dies can be replaced under the same record, with its apps rebuilt and its data put back. And quite a lot went into restores and moves being honest about themselves: Restore buttons that only said "coming soon", a notice that said a restore had succeeded when it hadn't, a progress bar that jumped back and forth. Small things, all of them, and exactly what a beta is for: finding them before someone's actual data depends on the restore.
What Didn't
The line didn't hold. The plan was to stop rewriting my own story and go out and tell it; instead I spent a good part of October rewriting it again. Not quite the discipline I was after, but I think it was the right call.
For a long time the pitch was getting your app live, which is what every tool in this space leads with. October showed pretty clearly what the job actually is: PHP, Caddy, a compromised certificate root, one advisory after the other, and every one of them patched, rolled out and checked on servers that belong to someone else, with a rollback ready in case the patch itself broke something. I'd been describing the easy half of it.
strackt is very good at getting your app live; that doesn't make it unique.
So the website moved from getting your app online to what happens after that. It took hours I'd promised to outreach; but refining the story and telling it right is just as important as the code that powers the platform, especially when it's your code, your data and your servers.
Would I Pay For This Today?
Yes, and more than last month. The new story for this website went out on it: a preview environment for every pull request while I iterated, then a merge to main and an automatic rollout, all on a fully patched ARM server of my own, and I didn't think about any of it once. It keeps getting more flexible; the more I test it, the more I use it, and the more I use it, the more I like it.
Next Month
November is about making all of it available to agents. By the end of it, an agent can onboard a server and deploy a repo from beginning to end. Telling an agent what you want is the easy half; the other half is security and recoverability, and strackt already has the checks, the steps and the rollbacks that make a change safe. November is about handing those over in a way that lets you ask for something and be confident about what happens next. Plus, hopefully, some good outreach; I've promised that before, so this time I'll just report back on how it went.
The story is finally clear enough to tell; now I have to go and tell it.