How to Run Demos

How to Run Demos at AI Tinkerers Meetups

How many?

For a 3-hour meetup:

• Max 8 demos, 5 minutes each
• Optimal: Two sets of 4 demos with a 10-minute break
• Acceptable: 6 or fewer demos in one session
• Interactive option: 4 demos with brief Q&A
• Prioritize open networking time

Goals:

1. Demos should be fun, technical, and inspire people to tinker.
2. A well-run demo section should be fast, lively, and result in connections between tinkerers. I hypothesize that the best way to facilitate that is to offer some imagination alongside the demo (basically, append a conversation starter to the demo) and advertise the presenters’ contact info.

Before the event:.

• Pick the speaker order ahead of time. There’s no rules to a great order, but I recommend starting and ending with your strongest presenters. (If you’re segmenting demos into two chunks, try starting and ending each chunk with strong demos.)
• Try to come up with something short to say about each demo, ideally something exciting that will encourage people to tinker more. The presenters are too humble to say the very cool, inspiring, futuristic implication of their tech, but you can! - Bonus points if you can find something that connects two presenters, or engages the presenter or audience (like a question). If you can’t come up with something to say for someone, that’s fine.
• Try to come up with a one-liner to describe the top ~5 sidelined demos. You’ll acknowledge them at the end of the event. Make sure you have the list of speakers and sidelined speakers ready to go on your computer/phone/paper.

At the beginning:

• Immediately ask the first presenter to start getting ready.
• Tell people about the time limit, 5 minutes max, say you will cut people off. (Optionally prepare some music to play at the end of the 5 minutes—this can be a kind way to cut someone off.)
• Tell people “the only company names should be in the stack trace” — this is not YC demo day, this is homebrew computer club. No slides, technical demos for a technical audience, and you must be RUNNING CODE.

For each demo:

• Set a 5 minute timer, but keep your eye on it.
• After 4 minutes, give the presenter the one minute warning (non-verbally if possible).
• After 5 minutes, play music if you’ve prepared it, or just walk up to the presenter.
• If you have only 6 to 8 demos total and the time is kept, you can say “we have time for two quick questions from the audience” - this allows things to be interactive and most valuable for everyone

At the end of each demo:

• Thank the presenter.
• Immediately ask the next presenter to come to the stage and start getting ready.
• Ask the last presenter how people can find them.
• Say something interesting about the last presenter if you can think of anything (ideally this was prepared earlier).

At the end of the demos:

• Acknowledge sidelined demos (ideally with some explosive one-liners that were prepared earlier) and say the names of the people. (Sideline acknowledgement is big—I’ve had people come up to me after the demos and ask for the contact info of sidelined presenters, who then went on to work together). Encourage them to apply again, and if there’s screens / “demo stations” available in the back, say they should demo there this time.

What about other formats besides demos?

Think about this: have you ever heard anyone ever say “I loved that panel discussion”? That is something people say to kiss up to a panelist they want to hit up for a job, but seriously, nobody has ever enjoyed a panel discussion. I’d outright avoid that format. It’s a lazy way to give airtime to “VIPs” but a terrible format for audiences – especially technical audiences like AI Tinkerers.

That said, it is possible to do panel discussions, fireside chats and spotlight topics. The critical factor is aligning and prepping the speakers: these must be highly technical, focused sessions covering exciting, novel and in-depth technical topics to be successful.

🔥🤖💻 Demo Stats:

5318 demos have been submitted and 3098 have been presented. The meetups have repeatedly showcased agentic automation, reliable AI via RAG/grounding and evaluation, and production-focused integration techniques such as standardized interfaces and orchestrated pipelines. Exciting technical themes have also included local-first efficiency, safety/guardrails, and multimodal/document/real-world domain deployments spanning robotics, healthcare, security, and analytics.

See also: What makes a great demo at AI Tinkerers?