Dear readers,

I went to check when I last sent one of these, and it turns out it was over two months ago. That surprised me — in my head it had been a few weeks.

I won't pretend that was the plan. The Good AI Guide is a side project, and side projects slip when actual work gets busy. But the gap did make me appreciate the people whose newsletters show up in my inbox every day without fail. Putting this site together over a weekend was the easy, fun part. Keeping it going week after week is the part that actually takes discipline, and I have a lot more respect for the people who manage it than I did before.

Anyway, thanks for sticking around. One thing that's kept me motivated: the site now gets around 5,000 unique visitors a month, and I've done absolutely no promotion beyond the occasional LinkedIn post. And I got new subscribers from all over the world. So at least some of you are finding this useful, which is genuinely nice to see.

I should also admit that I've been using AI in some way for almost all of my work lately and have learned a lot more than I've found the time to write up. There's a backlog, and this email is me starting to chip away at it. In the meantime, the site itself has changed quite a bit.

🎨 I rebuilt the whole thing

The site looks quite different now. I redesigned the homepage and the About page, and reworked how the guides themselves are read, all of it built with the same AI tools I write about here. That mattered to me. I wanted to see if the hype around the new Gemini 3.5 Flash in the updated Antigravity is justified (it's really, really fast, and quite smart).

The part you'll actually notice is the reading experience. Long articles are easier on the eyes now, there's a dark mode and a few reader controls, and the guides directory is finally organized in a way that makes sense. Take a look when you get a minute — I'm happy with how it came out.

🧭 I changed my mind about AI fluency

In the first email I told you about the AI Fluency Levels — five stages in a straight line, from "AI Curious" up to "AI Builder." Since then I've come to think that the straight line was wrong.

What I'd missed is that the path forks near the top. Some people go deep on strategy, thinking about how AI reshapes a team or a company or a whole way of working. Others go deep on building, actually making things, whether that's writing code, wiring up automations, or putting together a little app for a weekend trip. One isn't more advanced than the other; they're just different directions you can take. They are also not mutually exclusive.

So the model is now four levels — AI Curious, AI Explorer, AI Practitioner, and AI Advanced — and at the Advanced stage it splits into a Strategist lean and a Builder lean. I also reworked the scoring behind it, because where you actually stand has much less to do with knowing prompt tricks than with whether you treat AI as something you have a real back-and-forth with.

If you took the quiz last time, it's worth retaking — you might land somewhere different, and that's the interesting part.

✍️ Three honest pieces

These aren't how-to guides. They're more personal, the thinking-out-loud kind of writing, and they're the pieces I most wanted to get out of my head and onto the page:

• Why I'm Optimistic About the AI Future
  I get asked this constantly: is AI going to ruin everything? My honest answer — yes, the transition will hurt some people, and no, I haven't lost hope. Here's what I actually tell people who ask.

• The Orchestrator's High
  Coding agents now run on their own for ten or fifteen minutes at a time, and I can keep several going at once. The output is real. So is the racing brain, the skipped breaks, and the FOMO that fuels it. Notes from inside the loop — and how I'm trying to stay sane.

• Working With AI, Not Through It
  A few months ago I stopped using AI as a tool and started working with it as a partner. The shift didn't come from a better model — it came from the rules of engagement I keep rewriting around it.

📡 Who I follow, what I read

One more thing I never properly announced: I've started keeping two public lists on the site — the people I follow and the things I read. I read widely and filter hard, and these are the sources that actually survive that filter. If you're trying to build your own signal, this is a good place to start.

That's plenty for one email. There's a whole stack of more hands-on, build-it-yourself material I've been sitting on — how I shipped a custom app for a weekend trip, how I get coding agents to actually do real work, and a few tools that earned their place in my stack. I'll send that in the next edition, in a few weeks (hopefully).

As always, if any of this is useful to you, that's the whole reason I do it. Just hit reply if you want to tell me what's working or what you'd like me to dig into.

Best,
Fabian