I attended my first AI Tinkerers in-person event in Toronto. I'd been to their virtual paper club before and was duly impressed. Hot tip: the paper club runs at lunch, and if you consider yourself an AI enthusiast (are you even on LinkedIn if you're not?), join in for regular AI enlightenment.
What I didn't realize until the in-person event was just how much energy and joy fuels this community. I was a little nervous going in. Some projects went over my head, and I don't yet have the technical depth to keep up with everything. It didn't matter. This group is warm, inclusive, and encouraging. They actually get excited (instead of judgmental) when you tell them you vibe code, which feels like the natural reaction to me but apparently isn't the norm in the wider world. I'm glad to be in this one instead.
That reaction deserves a paragraph of its own, because it's diagnostic. You can tell a lot about a technical community by how it treats people at the edge of their depth. Gatekeeping communities treat accessibility tools as cheating; generative ones treat them as more people to build with. The AI Tinkerers crowd is unambiguously the second kind, and it's not a coincidence that the second kind is where the interesting projects show up. Curiosity compounds when nobody's afraid of asking the dumb question.
The projects, and what each one taught me
OPEA — a framework for deploying agents in enterprise environments. Deeply technical, mostly beyond me, and I loved it precisely because it gave me homework. Enterprise agent deployment is exactly where my product work is heading; knowing what I don't know yet is half the value of a night like this.
GetKiln.ai — a rapid fine-tuning and evaluation platform that helps you figure out which cheaper model can do the job just as well. I watched the presenter spend many dollars very quickly, and was impressed. This one is directly relevant to anyone shipping AI products: "which model is good enough?" is a product question wearing an engineering costume, and tooling that answers it fast changes how you scope MVPs. On my list to experiment with.
"Dream" engineering — a creative showcase of how tech can empower people who think they aren't creative but, in the speaker's words, just aren't being honest with themselves. The framing stuck with me more than the tech.
An open-source meeting bot — I assumed this would be boring (meeting bots exist, right?). Wrong. A group of undergrads built an open-source alternative because everything on the market is expensive, then containerized the whole thing in Terraform so developers can actually deploy it instead of fighting 25 AWS setup steps. Technical elegance in service of accessibility, presented with infectious energy. Personal reminder: don't judge a project by what it sounds like.
VAEs vs JEPA — fine-tuning computer vision with JEPA on a personal GPU, knowingly more expensive than the standard approach, purely for the learning. There's practical entertainment value in seemingly impractical research, and honestly, more of us should budget for it.
A live prompt printer — the most popular one-liner of the night, printed as an image, on a custom-coded printer server. Completely unnecessary. Completely delightful. A high note to end on.
The science fair, and the subway coder
After the talks came a science fair of projects. I only managed one real conversation, with someone who vibe codes projects on the subway on his way to work. On the subway.
It connected perfectly with a conversation I'd had recently in the AI Women's Lounge: AI is going to give us more time, and what we choose to do with it is entirely up to us.
Why practitioner rooms beat courses
Here's the thing I keep coming back to. I could have spent that evening on a structured AI course, and on paper the course would have "taught" me more. But courses teach you the consensus, and consensus in AI is roughly four months old at any given moment. A room full of practitioners shows you the frontier: what people are actually building, what's breaking, what's suddenly cheap. One evening recalibrated my sense of what's possible more than a month of reading did.
Watching these talks live also made me more excited to one day present something of my own. That's the other thing good communities do: they don't just inform you, they raise your ambition. I have a working app and a growing pile of opinions about building AI products in regulated industries. One of these days, that's a talk.