You train models, wire up pipelines, read papers way past bedtime and sometimes lose an argument with a loss curve. Once a week, people who do the same thing get together in one room in Taipei to compare notes out loud.
How a session runs
while week:
lightning_talk() # if someone has one
open_discussion()
maybe(demo, deep_dive)
Here's what that loop means. If someone has a talk ready, we start with a short lightning talk on one technical AI topic. It's meant to be concrete enough to argue about, not a keynote. After that the group takes over. People share approaches, ask questions, bring up experiments that worked and ones that didn't, and the conversation goes wherever the room wants.
Some weeks it turns into a group deep-dive or a live demo. Other weeks it's a long, useful argument about one design choice. Both count.
What you'll find in the room
A tight opener: one short lightning talk to get things going, when there is one
Discussion led by the group: technical and focused, with no slides required to join in
Hands-on moments now and then: joint deep-dives and demos
People who build: builders, not people who only post about it
Who it's for
AI engineers, researchers and hands-on builders at any experience level. You don't need a job title or a publication record. Bring an interest in how things actually work and a willingness to think it through with others.
Bring a question you're stuck on, an experiment that half-worked, or an assumption you'd like other people to poke holes in. You'll leave with sharper instincts and a few more people to text when your eval numbers look suspiciously good.