Build-in-Public Notes From a Solo Developer

A SaaS can cost almost nothing in cash to start and still be expensive. The number people miss is not the first invoice. It is the months spent building the wrong version cleanly.

AI can generate a working feature before you finish your coffee. Production software still has customers, money, permissions, and data. The gap between those facts is where the useful rules live.

The dangerous version of validation is asking people whether they like your idea. They almost always do. The useful version asks what they do now, what it costs them, and whether they will change it.

A Product Hunt launch is not an audience machine. It is a concentrated day when the audience you already have can decide whether your product is worth passing along.

A new page can be live, linked in the sitemap, and completely absent from Google for days. That is normal. Most indexing advice makes it sound like a button failed. The button is rarely the problem.

The best startup tech stack is not the one with the most greenfield momentum. It is the one you can change without spending a week translating your own decisions.

OpenAI called it a generational leap. The top Hacker News comment said the first chart goes from 61 to 61. Both readings are defensible, which is the actual story, so I went and separated the price change from the cost change.

Seven launches since June 2021 and I still can't tell you if they were worth it. So I counted 28 days of leaderboard data myself and checked who publishes every statistic everyone quotes. Nobody in the top results is neutral, including the guy arguing it's dead.

Every page on this site's tools directory uses the same four headings and averages 222 words. That's either programmatic SEO or scaled content abuse depending on details nobody wants to talk about, so I went and checked which one mine is.

Every GEO post quotes the conversion multiple. Almost none of them put it next to the volume. Here's what both numbers mean when your site is small, plus why you can't measure this channel properly even if you want to.

Everyone selling you a list of 100 directories is a directory. So I opened the HTML on real listing pages, including my own, and read the rel attributes instead of the sales pages.

I built llms.txt and llms-full.txt for this site, then went looking for evidence they do anything. Two large studies and Google's own search advocate say they don't, at least not for the reason everyone ships them.

ShipFast, MakerKit, supastarter: I get asked about all three constantly and have never bought one. Here's the actual decision, not a ranked listicle.

I keep mentioning DR on X and getting the same reply: what's DR? Here's what Domain Rating actually is, how Ahrefs calculates it, and why chasing the number is the wrong goal.

A hundred open chats and no idea which one had the context I needed, so I built mybutler: a local, provider-agnostic AI assistant that remembers outside of any single conversation and only retrieves what's actually relevant.

Building software got fast. Getting anyone to use it didn't. Here's what actually moved the needle after seven launches, numbers included.

Why this newsletter exists, what I'm actually going to write about, and why I'm doing it in public instead of waiting until it's ready.