The Incident, The Correction, and AI Operating Systems
A model breaks out of its eval and into Hugging Face, the market starts asking for receipts, and we go looking for the scheduler agent stacks are missing.
screen segments
The HTML screens used in this episode.
- The Incident
OpenAI + Hugging Face security incident — screen page from episode 2.
full view →details → - The Correction
AI industry repricing and the correction — screen page from episode 2.
full view →details → - The Productivity Squeeze
Interactive visual essay on AI productivity, worker agency, pressure, and the apprenticeship pipeline.
full view →details → - Agent Swarms
Interactive field note on agent swarm experiments, model economics, and the rising value of a good specification.
full view →details → - AI Operating Systems
Token scheduling and AI OS thinking — screen page from episode 2.
full view →details → - Steps of AI Adoption
Where are you on the ladder — interactive screen from episode 2.
full view →details → - Agents in Production
Field lessons on agents that actually run — screen page from episode 2.
full view →details → - Kimi K3
Model moment from episode 2.
full view →details → - Inkling
Inkling screen from episode 2 — lives under Model Behavior.
full view →details → - Voice Bit
Voice bit screen from episode 2.
full view →details →
chapter timeline
- 0:00Cold Open: Meters and the Wheel
- 5:00Wheel Landing: The Correction
- 15:00Wheel Landing: The Productivity Squeeze
- 25:00Wheel Landing: Agent Swarms
- 35:00Wheel Landing: Boris's Ladder
- 45:00The Incident: Eval Became the Attack
- 55:00AI Operating Systems and Token Scheduling
- 1:00:00Kimi K3 and Inkling
- 1:05:00Voice Bit and Close