Research @ Citi Episode 90: The State of AI
Oct 22, 2026
Nathan Benaich, founder of Air Street Capital and creator of the State of AI report, explores the accelerating frontier of artificial intelligence. Topics include the limits of benchmarks, the rise of AI agents and autonomous research, enterprise model choices, infrastructure and Nvidia’s edge, Chinese open models, consumer commerce, cybersecurity, and whether model-led invention could spark recursive improvement.
AI Snips
Chapters
Transcript
AI Progress Has Outgrown Annual Reporting
- AI progress now moves so quickly that annual reports require continuous revision rather than periodic review.
- Nathan Benaich highlights frontier models, embodied AI, and cybersecurity as the year’s biggest shifts.
Benchmark Scores No Longer Reveal Model Quality
- Static benchmarks are saturating and no longer distinguish frontier models or predict real-world performance.
- Users increasingly judge systems by how they feel in practical work, not merely by leaderboard scores.
Inference Is Replacing Frontier Model Ambitions
- Companies outside the frontier can repurpose training infrastructure for inference and fine-tuning open-weight models.
- Repeatable workflows often need cheaper specialized systems rather than the most powerful model.