Posts will appear here as I write them.
Posts
-
Using Rift Stats live at a tournament, and where it fell short
Today was a tournament day, which meant less time building and more time actually using what I’ve built. That’s a useful kind of test.
-
Rift Stats becomes a real tool: live event querying and what that means
A productive couple of days, and ones where the project shifted in a direction I didn’t fully anticipate going in.
-
Codex frustrations and the tool invocation problem
Today was mostly friction, which is useful in its own way.
-
Catching up: match history architecture and finding easy tournaments
A few days lighter than usual — some illness and weekend commitments — but not completely idle. I was able to push a few features forward on projects over the weekend without much active intervention, which is its own kind of useful data point about how far along these projects are.
-
Building Gym OS: real architecture, real DevOps, real approval fatigue
Today was the first day that felt less like tinkering and more like building something with actual scope. I’ve been calling it Gym OS — software for managing a gym, a domain I know well enough to have real opinions about. I had a plan drafted from the night before and spent the day executing on it.
-
Cross-agent collaboration and unsupervised loops
Today was mostly experimental — less about shipping something and more about understanding what different workflows feel like in practice.
-
Remote Claude sessions, Railway hiccups, and a new project spec
Today had a mix of progress and friction, which is probably the most realistic description of how most of these sessions go.
-
AI slop UI, Claude skills, and finding the right feedback loop
A short but productive day. I made a few small bug fixes on the PWA through the mobile workflow, and then switched focus to the frontend quality problem I’d been thinking about since yesterday.
-
Shipping from my phone with Railway and PWAs
Today started with a question I’ve been putting off: what does DevOps actually look like for a solo developer with limited resources? I wanted to understand the reasoning behind the decisions Claude made in the initial Riftbound stats frontend, think through continuous deployment, and figure out how to make development and testing work from a phone. By the end of the day I had a working answer to at least part of that.
-
Data pipelines, a committed .env, and a new writing tool
I took yesterday off sick, so today was about getting back into it. Most of the session went into the Riftbound stats project — a card game data analysis side project I’m running in parallel with the blog — but I also set up something new for the writing workflow itself.
-
Solo-building with AI
This site exists because I want somewhere to think out loud about what I’m learning. That’ll be a mix of things — building software with AI agents, general observations about AI, and applying ideas from other places I’ve encountered them. I’ve used AI in a professional context before, but this is different: figuring out what’s possible as a single person, with no team or existing codebase to lean on. Today was the first real session and it was a good reminder of why critical thinking matters even when (especially when) you’re delegating to an AI.
subscribe via RSS