AI·Signal

AI Signal — 2026-11-19

Status update — no expert publications

No tracked expert published on 2026-11-19. Discovery ran across all 10 monitored channels and found nothing new, so no synthesis was produced rather than manufacturing analysis from an empty source set.

Standing Assessment · carried forward from 2026-11-18

Anthropic's S1 (about $81B run-rate on under $10B net burn, ~85% paid-inference gross margin) shows that frontier AI at scale makes money, and that the money comes from agentic coding. Model leadership turns over about every six months, open-weight tokens sell at 10-20% of lab prices, and vendor pricing (Gemini 4's expiring discounts, a tier dropped without notice, effort-level multipliers) is getting harder to read. Value is moving toward application-layer agents, harnesses and routing, while raw tokens lose pricing power. Enterprise diffusion is the main constraint, not model capability: even $100K+ accounts spend under 0.1% of opex on tokens. Security has not kept pace with agent permissions.

Panel Activity

ExpertLast publishedGap
Daniel Miessler2026-11-109d earlier
Nate B. Jones2026-11-181d earlier
AI Explained2026-11-109d earlier
Dwarkesh Patel2026-11-181d earlier
Matthew Berman2026-11-163d earlier
Latent Space2026-11-181d earlier
Simon Willison2026-11-181d earlier
Hamel Husain2026-10-2030d earlier
Nathan Lambert2026-11-145d earlier
SemiAnalysis2026-11-127d earlier