I Over-Engineer Everything. A Room Full of Claude Users Set Me Straight.

I hosted a Claude Users Group meetup in Pasadena last week, and I’ll be honest: I spent the days leading up to it quietly convinced that four people would show up, and three of them would be there for the free pizza. Just over thirty folks walked through the door at FoundrSpace on Raymond. For an unofficial, no-pitch, “come talk about how you use Claude” gathering that I put together mostly on a hunch, that felt like a win.

The premise was simple, and I kept it that way on purpose. Fellow Claude users in a room. Roundtable format. Ask questions, swap ideas, and, if you want, take 10 minutes to show the rest of us something you’ve built or figured out. No product pitches, no vendor booths, no “and if you sign up today.” Just people who use the tool sharing what they actually do with it. Anthropic had nothing to do with it, to be clear. This was a bunch of us in a coworking space with coffee.

Overengineering, a bug not a feature

I over-engineer everything. It’s a character flaw. You hand me a problem that needs a Post-it note, and I’ll hand you back a plugin, a webhook, and a dashboard nobody asked for. So the part of the night that stuck with me wasn’t the clever technical stuff. It was watching people use Claude in ways that were almost aggressively reasonable. No cleverness for its own sake. Just tools being used the way they’re meant to be used, on real problems, by people who had zero interest in showing off.

That’s something I need a reminder about, honestly. I’m the guy who builds little custom tools for problems that probably didn’t need them. Watching a room full of people just quietly getting things done was a good corrective.

It’s really easy to build a messy solution to a fussy problem. But you know what takes the cake? Restraint. Reading the fucking manual (RTFM!) and doing what the tool prescribes. Those people who do that? Rockstars all of them. I don’t have restraint; let me at the tool and bonk it around until something useful comes out of it. It’s pure chaos how I approach learning and building.

HVAC use case and the proposal templates

One of our attendees, Rosalia, shared how she uses Skills in the Claude desktop app to templatize her husband’s proposal writing. That’s it. That’s the whole thing. Her husband writes proposals, the structure is mostly the same every time, so she built a Skill that captures that structure and does the repetitive lifting.

And I love it precisely because it isn’t fancy. There’s no grand system behind it. She found a task that was annoying and repetitive, and she used the tool as it’s designed to be used to make it less annoying. That’s the good stuff. Half the value people leave on the table with these tools is waiting for a big, impressive use case, while the actual wins are small, boring, and repeatable. Rosalia didn’t wait. She just did the thing.

Noel’s second brain for wall-of-text replies

Noel Saw, a UX designer who runs his own agency, shared something I hadn’t seen anyone do before. You know how Claude will sometimes answer a simple question with a beautiful, exhaustive wall of text, and you’re left going “okay, but which part do I actually act on”? Noel built a set of Skills, basically a second-brain database, that steps through those conversations logically and breaks that wall down into something workable.

So instead of drowning in a giant response, he’s got a repeatable way to make the model organize itself into next steps. As a UX person, of course he did. That’s a UX problem applied to an AI conversation, and it’s genuinely innovative. I’ve been chewing on it ever since.

The best uses had nothing to do with my job

Here’s what I keep coming back to. I live in web development. When I reach for Claude, I’m usually reaching for Claude Code or something knee-deep in a build. But the uses that lit me up at this meetup had almost nothing to do with any of that. Proposal writing. Taming a conversation. People solving the friction that shows up in their day, in their work, which looks nothing like mine.

That’s the whole promise of these tools, and it’s easy to lose sight of it when you spend all day using them for one narrow thing. The people getting the most out of Claude aren’t necessarily the most technical people in the room. They’re the ones who spotted a repetitive task and thought, “I bet I can hand this off.” It’s the same instinct behind the automations that quietly save us a day a week and the reason I keep telling people they’re underutilizing the tools they already pay for. The wins are hiding in the boring stuff.

If you want the more technical end of that spectrum, I’ve written about using AI and automation on actual client work too. But that’s not what made this night good. What made it good was the range.

We’re doing this monthly

The turnout and the energy made the decision easy: I’m running this monthly, in Pasadena and in Chinatown. Same format. Show up, share what you’re doing, learn from someone whose work looks nothing like yours. No pitches, no gatekeeping, no prerequisite level of technical skill. If you use Claude, or you’re curious about it, you belong in the room.

Big thanks to everyone who came out, and to Rosalia and Noel for standing up and showing us their stuff. If you were there and I didn’t get to your use case, come to the next one and corner me. And if you missed it, now you know it exists. Come say hi. I promise not to over-engineer the coffee.