What a strong AI sponsor track looks like
Lessons from AI-sponsored weekends (including Claude Speedrun): briefs builders can finish, demos judges can score, prizes that do not warp the room.
July 27, 2026
Logo slides do not ship demos
When Anthropic-backed energy showed up at Claude Speedrun (~$10k prizes, ~80 hackers), the useful part was not the logo wall — it was a problem space narrow enough to finish in a weekend.
Write a brief, not a moodboard
A strong AI track answers:
- What problem should teams attack?
- What tools are in-bounds (APIs, models, data)?
- What is out of scope?
- What does a winning demo show in ninety seconds?
"Build something cool with AI" produces slideware. "Extract structured claims from these PDFs and show provenance" produces comparable projects.
Judging criteria before doors open
Publish the rubric with the brief. Axes that work:
- Correctness / reliability
- Quality of evaluation
- Product clarity for a real user
- Technical ambition only after the first three are not broken
If "used our model" is the top criterion, you will get wrappers.
Prizes shape behavior
Cash still talks. Credits are fine as secondary prizes. Interview fast-tracks belong in the FAQ. Public ballparks from our past events live on Previously — do not invent new numbers for LinkedIn.
Company challenges on our stack
Sponsors who want apply → submit → judge → gallery without duct-taping Forms to a chat app: company challenges. Student organizers: host free with .edu.