The Best Part of AI Coding Wasn't the AI

On July 4, I spent the day at Cafe Cursor Beijing. I went expecting to talk about AI tools. I left thinking more about the people using them.

The event took over a bright cafe and filled it with laptops, half-finished products, and developers who were eager to compare notes. Even the drinks menu was written like an API response: the server was Beijing, the service was Cafe Cursor, and the model was Composer 2.5.

It was playful, but it captured the atmosphere well. Everyone had arrived with the same broad interest in AI, yet no two people were using it in quite the same way.

The Conversations Mattered Most

I met developers working on different products and at different stages. We exchanged views on where AI is genuinely useful, where it remains unreliable, and how it is changing the way we approach everyday development. We also shared what we were currently building and the problems occupying our attention.

Those conversations were more valuable than a list of features or another polished demo. Online, discussions about AI often collapse into extremes: it will replace everyone, or it is nothing more than autocomplete. In person, the conversation was more practical. People talked about what had worked, what had failed, and what still required human judgment.

That groundedness was refreshing.

Faster Execution Still Needs Better Judgment

Earlier this year, I wrote about the illusion of certainty in AI coding tools. My conclusion was that AI should accelerate your thinking, not replace it.

Cafe Cursor did not change that belief. It made it more concrete.

AI can shorten the distance between an idea and a prototype. It can remove repetitive work and help a small team attempt something that previously required far more time. But choosing the right problem, evaluating an answer, and explaining why a product should exist are still human responsibilities.

The interesting part was not that everyone had access to the same tool. It was that everyone brought a different set of experiences, priorities, and questions to it.

Why the Room Mattered

This matters to me beyond a single event. As a co-founder of Cobay, I help organize offline founder meetups while also building the website and product behind the community.

Cafe Cursor showed me what a good physical space can do. Put builders in the same room, give them enough time to talk honestly, and ideas begin moving between people. A technical problem becomes a conversation. A vague project becomes something another person can question, sharpen, or support.

A community is not created by putting people into a group chat. It is created through repeated moments where people feel comfortable sharing unfinished work.

That was the best part of Cafe Cursor Beijing. The AI brought us into the room, but the people made the day worth remembering.

The cat, admittedly, also helped.