I locked 1000 architects in a room until they all agreed on the best solution
summary
Generates 1,000 personas (names, photos, specialties) and 1,000 architecture proposals from LLMs, uses DAL‑E/GPT image tools and reasoning models (04‑mini/O‑series), then runs a tournament‑style knockout where proposals are blindly scored in JSON by peers (4→7→998 reviewers) to pick a winner; finds bias from persona metadata, cost/token tradeoffs (50M tokens used), surprising regional privacy tailoring, and outputs rich mermaid diagrams and deployable designs.
AI-written summary · about
chapters (8)
- 0:00AI-driven architecture experiment
- 3:58Consulting AI architects
- 13:19Diverse architect personas
- 26:03Tournament to evaluate architectures
- 33:28Running the tournament
- 42:05Key takeaways
- 51:55AI-driven architecture reviews
- 56:05Q&A
related talks
48m(Unpacking) AI's Impact on System Architecture
16mEveryone architect!
39mOne Does Not Simply Make Architecture Decisions Alone: ADRs and Architec
01h 02mThinking Like an Architect
56mEssential Architectural Thinking – Why, What, How, When and How Much?
39mA commune in the ivory tower: A new approach to architecture
45mThe AI Software Architect: Revolution or Illusion?
40mThe (too) many faces of architecture