How to talk to statues — Joe Reeve, ElevenLabs

TL;DR

• 1

  A two-hour prototype triggered real business demand: Joe Reeve built the statue app in Cursor on a Sunday, posted it with a video, and watched it jump from 50,000 impressions to 1.5 million while museums, Bonhams, Christie's, and travel companies reached out.

• 2

  The app works by chaining existing APIs, not inventing new core tech: a photo of a statue triggers OpenAI deep research, ElevenLabs voice design, and an ElevenLabs agent, producing a voice call with the object in about 30 seconds.

• 3

  Museums care less about the demo and more about narrative quality: Joe says the long-tail problem is not scaling infra but getting curators to define what a statue should say, instead of relying on whatever a model pulls from Google.

• 4

  Voice UX is still awkward because people are too polite to machines: the group kept coming back to a core friction where users hesitate to interrupt agents, do not know how long an answer will be, and want multimodal responses like text, diagrams, or visual summaries alongside speech.

• 5

  Vibe coding is widening who gets to make software: Joe compared current vibe coding meetups to early hackathons, with people showing up who have never thought about how apps are made and asking for interfaces in completely non-engineer terms.

• 6

  Short-form video craft mattered as much as the code: Joe edited the viral statue clip on his phone in about 20 to 25 minutes using CapCut, stressing fast hooks, captions, and music as the practical ingredients that make people actually watch.