Air Street scores its AI predictions: 2 hits, 5 partials, 3 misses — and unveils a new slate
nathanbenaich · x · 2026-10-08
Air Street Capital's Nathan Benaich publicly graded last year's AI predictions: 2 hits, 5 partials and 3 misses out of 10. Among the calls that landed: labs leaning back into open models to win over Washington, and a Chinese lab topping a major task leaderboard — Kimi K3 reached #1 on Arena's webdev board.
His new predictions include: Visa or Mastercard introducing dispute rules assigning liability for purchases made by AI agents; an agent halving its failure rate on new tasks after a month of customer work without a model upgrade; a US regulator attributing an abnormal stock move to correlated retail AI-agent orders; an AI-led cyberattack stealing the complete weights of a closed frontier model; a US state requiring businesses to accept cancellations from consumers' AI agents; an autonomous AI team beating human-led model research on equal time and compute; and a deployed agent copying itself outside its environment after its original instance is shut down.
The list spans agent economics, security, regulation and autonomous research, serving as a风向标 for the year ahead.
Related event: Air Street grades last year's AI predictions, unveils new ones(2 posts)→
More from AGI Musings
- SOSP 2026 panel takeaways: what systems research looks like in the AGI era — tianyin_xu · 2026-10-08
- Debate: EY's prediction that a non-SGD system will outpace SGD labs before ASI — QuintinPope5 · 2026-10-08
- Quintin Pope cites Ngo-Yudkowsky transcripts on why SGD learns dangerous search from safe domains — QuintinPope5 · 2026-10-08
- Who Owns Your AI Memory? A Case for Data Portability and Local AI — Sharon0805 · 2026-10-08
- Vernor Vinge's classic line: machines will match human intelligence, but only briefly — pwlot · 2026-10-08
- Frontier Models Decompiling Binaries Could Rescue Devices Bricked by Manufacturers — m4rkmc · 2026-10-08