On 15 Oct 2026 I posted on X that 80% of software & AI stocks are down today. The same day I also noted that five frontier AI models are within five intelligence points of each other, yet their cost per unit differs by 3.8×.
80% of software & AI stocks are down today.
Five frontier models now sit within five points of each other in terms of intelligence, but their cost per unit differs by 3.8x.
I clarified in a reply that the appropriate metric is cost per task, not per token, to correct a misunderstanding in a prior discussion with @thomasjankowski.
- Open questions:
- What benchmark defines the "five points of intelligence"?
- Which five frontier models are compared?
- How is "cost per unit" calculated?
- What drives the 3.8× cost disparity?
- Market implications of the 80% stock decline?