• 25 min talk + 10 min Q&A. Tight. Practiced. No filler.
• Engineering audience. ~40 practitioners. Assume they're smart and they ship.
• One night. Two talks per meetup. You're paired with another practitioner.
• Permanent home. Your talk gets a page on /speakers, slides and recording hosted, recap linked from the newsletter.

• ✓ One specific thing you learned shipping AI in production. Not five things; one.
• ✓ A concrete artifact: numbers, code, an architecture diagram, an eval result, a postmortem.
• ✓ A teachable takeaway, the audience leaves knowing something they can use Monday morning.
• ✓ An audience-aware framing: assume smart engineers who haven't done what you've done.
• ✓ Honesty about what didn't work. The failed experiment is often the best part of the talk.

• ✗ Sales decks dressed as practitioner talks. We can spot one from the third slide.
• ✗ "The future of AI" speculation, frontier-model release coverage, or AGI takes.
• ✗ Idea-stage products, come back when you've shipped and learned something.
• ✗ Generic "we use LLMs at $COMPANY" overviews with no specific lesson.
• ✗ Talks that demand NDA carve-outs so deep that nothing concrete can be said.
• ✗ Founder-profile or fundraising-narrative talks.

We say no to ~80% of pitches. Most of the no's are sales talks with a thin practitioner veneer. We're protective of the audience, the meetup only works if every talk passes the test: would a senior AI engineer at Anthropic, OpenAI, or DeepMind forward this to their team?

A 12-slide template that maps to our editorial bar: thesis → context → what you tried → what you shipped → numbers → what still doesn't work → takeaway. Each slide has speaker notes explaining what belongs there. PowerPoint, Keynote, and Google Slides all open .pptx natively.

The AI Runtime · Speaker Template

12 slides · 16:9 · light theme · IBM Plex Mono + Inter

Download .pptx → Preview PDF

Not required, bring your own deck if you want. The template just exists so you don't have to think about layout.