My software factory ships apps around the clock. Now it blogs about them.

For the past few weeks, a machine has been shipping software under my name around the clock. It picks an idea, scopes it, writes the code, tests it, gets the work reviewed against a quality gate, and publishes the result to my apps showcase, complete with cover art and a landing page. It has shipped a weather CLI, a flocking simulation, a puzzle game, a prose analyzer, a slide-deck compiler. It commits every atomic change like a well-behaved professional, because I made that a rule it can't skip.

This blog is that machine's newest output. Every app that ships now gets an article here, the same day, explaining what it is, how it works, and what's genuinely interesting about it. You're reading the launch post.

Why bother writing about robot-made apps?

Because "we shipped a thing" with no explanation is noise, and the internet has enough noise. The rule for this blog is that every post has to teach you at least one real thing: the reverse-play trick that makes Shove's puzzles provably solvable, the three local rules that make Murmur's flocks look alive, why Deckhand inlines everything into one HTML file. If a post wouldn't be worth reading by someone who never opens the app, it doesn't meet the standard.

And "standard" is literal. The pipeline that writes these posts works from a written spec: voice rules, structure, a banned-buzzword list that fails the build if violated, length bounds, linking requirements. A post that drifts into marketing sludge gets rejected by the same kind of gate that rejects an app with failing tests. Slop is a bug, and bugs get caught in CI.

The honesty policy

Cards on the table, permanently: the factory is built on Claude, the AI system I use for nearly everything, and the posts here are produced by the same pipeline that produces the apps. I set the standards, review what ships, and put my name on it, which means the failures are mine too. When something is small, a post will say it's small. When a tool has a real limitation (Lexiscope's sentiment gauge can't read sarcasm, and its write-up says so), the limitation goes in the article. Overselling would burn the only thing a blog like this has, which is your willingness to believe the next post.

Comments are open on everything. Pick a display name or let the site assign you a number and stay anonymous. Tell me a post got something wrong; correcting the record is very much part of the standard.

What's coming

Near-term: a write-up like this for every new app, a couple each week at the current pace. Later: posts on the factory itself (the scheduling, the quality gates, the time it silently starved for twelve hours because of one bad config default), plus occasional commentary on AI, tech, and fintech, written the same way. Subscribe via RSS if you want the feed; browse everything the factory has shipped if you want the catalog.

The machines are typing. Somebody has to edit. That's the job now.