What stood out wasn’t the model talk but the downstream assumptions: that AI systems will be cheap, reliable, and predictable enough to sit inside robotics, manufacturing, and infrastructure loops.
The deck mostly treats those as solved or solvable via scale.
If you’ve worked on real‑world AI systems, the gaps between those assumptions and current failure modes are where the discussion gets interesting.
The deck mostly treats those as solved or solvable via scale.
If you’ve worked on real‑world AI systems, the gaps between those assumptions and current failure modes are where the discussion gets interesting.