Can You Keep a Stadium? Make Them Feel Smart.
summary
A concrete claim: portable user memory keeps data off-model and switchable between agents, enabling customers to post-train sovereign models and deploy them in their own cloud or airgapped Macs. The stack compresses large models via quantization, runs a 8B model on a MacBook Air, and passes the Locomo benchmark at 100% on an 8B instance. Example: a 73‑year‑old nontechnical parent moves from “AI is not for me” to planning trips, buying a drone and getting licensed after a podcast plus a single ancestry chatbot workflow.
AI-written summary · about
chapters (17)
- 0:00Introduction and context
- 1:12Stadium analogy for pitching
- 2:36Why people reject AI
- 4:01Enterprise AI deployment failures
- 6:07Engaging my dad with AI
- 8:14Clarity and kindness in explanations
- 9:21AI adoption survey results
- 10:45Introducing Backboard’s approach
- 11:27Portable memory and data control
- 12:53Model portability and compression
- 13:56Quantization explained
- 15:00Air-gapped AI on a MacBook
- 15:44Sovereign AI vision
- 16:27Do people need AI?
- 17:32Reaching non-technical audiences
- 19:41Overcoming philosophical resistance
- 21:50Q&A summary
