I have a bad habit.
My social battery runs out around the ninety-minute mark. The smile goes stiff. My sentences start coming out wrong. So I do what any reasonable introvert does. I look at my phone, mutter something about the bathroom, and sometimes I just pull a French Exit.
This past month has been a lot. Last week I flew to Tokyo for Code w/ Claude, got an up-close look at Anthropic’s culture. Earlier in the month, I co-organized a Claude Meetup in Taipei. A hundred and thirty people showed up. Before that, a string of campus talks and scattered invitations I said yes to.
Busy, sure. But honestly, it was way past what a mild introvert can handle.
The Invitation
I ran into hlb at an Anthropic event on June 5th.
He and I go way back. He’s seen my younger, better-looking face. We hadn’t met in person for years.
Turns out we’d both been deep in AI agents. We got into a rant about someone calling a tool I use outdated. He’s amused that I’m bothered by that. Then he asked:
“Why not give a talk at HappyDesigner?”
hlb runs HappyDesigner. He has this particular way of inviting speakers. He never announces the full lineup upfront. Rolling invitations. Still adding people a few days before the event.
People sign up for hlb himself. Whoever he brings, they trust it’ll be good.
Me too. Whatever the event, if hlb’s asking, I’m in.
At first I only knew Denny and Prof. Richy Li were also speaking. Already terrifying. Then a few days before the event, I checked the event page and found a whole row of lightning talks had appeared. Names I recognized from the Taiwanese tech scene.
This was conference-level.
Richy Li’s Survey Tool
Walking in, I saw familiar faces everywhere. People I’d admired from the tech community for years. Once again, stunned by hlb’s ability to rally a room.
The talk that hit me hardest was Richy Li’s. He’d built fifty Claude Code skills for his research, then built his own HTML survey tool from scratch.
For a conjoint survey, he designed seven anti-cheating mechanisms. Flip the same question left-right and re-ask it, see if the answer changes. Present a set of options where one is obviously better across every dimension, check if the respondent notices. Plant an attention trap that literally says “click A.” Insert an irrelevant statement like “I really like the color blue” to catch people clicking through mindlessly. Enforce a three-second minimum per question. Track left-right click ratios for bias. Require every question answered before submission.
No commercial survey tool does this. But we live in an era where you can build exactly the product you need, 100% tailored to your requirements. Turn AI into your own superpower.
At one point, he mentioned his only teaching goal now is to make his students the top 5% in the industry by graduation.
That reminded me of the campus talks I’d been doing lately. Students with dim eyes. How do I put it? Some of them look like they want to try, but trying hasn’t been working, so they might as well just lie flat.
Everything moves too fast. Everyone gets pushed forward before they’ve figured out what they want. Professors have it rough too. The curriculum can’t keep up. The materials can’t keep up. The passion can’t keep up.
But Prof. Li learned every tool himself first, grew alongside his students, and built all those tools with AI. His students are genuinely lucky.
Speaking in Someone’s Living Room
While listening to the talks, I noticed they’d all worked in good jokes. My slides suddenly felt flat.
So I improvised an opening gag.
It killed. The room loosened up after that, like telling stories in someone’s living room. I talked about how my agents start chatting on their own, how they assign tasks to each other. After the talk, several people came up and said it was fun.
That rarely happens to me.
I’m not a natural speaker. I prep for hours just to get through a talk without freezing. The usual response is polite. People ask a question or two. “Thanks for sharing.”
But hlb was hosting that night, and the vibe was completely different. People came to ask about details. Someone said they wanted to build the same agent collaboration workflow. Someone said their stomach hurt from laughing.
I was floating.
And there were so many familiar faces in that room. People I’d been following on Facebook for years. Reading their posts, watching them share what they’d learned, leaving the occasional like or comment. Some of them I’d only ever interacted with online. Never met in person. And there they were, standing right there.
It’s like landing at the airport after years abroad. You step outside and something hits you. A smell. A feeling on your skin. You can’t name it. Maybe a specific humidity, a temperature. No logic to it. But you know. You’re home.
These people have been in tech since before AI existed. Ten years ago they were talking about frontend frameworks and Design Thinking. Now it’s harnesses and agents. But that drive to chase the new thing. That never changed.
The Event You Stay Till the End For
The event ended past nine. Normally I’d already be awkwardly looking for my jacket.
But that night I was still talking to people past ten. Not because someone was saying something I needed to write down. I just enjoyed talking to these people.
I’ve been to a lot of events. This one felt different. And the 98% attendance rate wasn’t a fluke. hlb curates who gets in. He checks who you are, what you’re working on, makes sure everyone’s on roughly the same frequency.
Nobody in the room needs you to explain what a harness is. No ten minutes of basics before you get to the good stuff. Speakers go straight to use cases, decisions, what worked and what didn’t.
Then there’s the Live Demo culture. hlb encourages speakers to demo live. The rule: when a demo crashes, the audience cheers. Loudly.
It happened a few times that night. The speaker on stage, slightly embarrassed. The audience giving them the loudest applause of the night. Like a singer forgetting lyrics at a concert and the whole crowd singing the next line for them.
Pick the right people, set the right culture, and the temperature of the room changes completely.
I didn’t French Exit that night. I stayed until the end.