Samaya AI beat Wall Street consensus on 40 of 50 Q3 earnings predictions

maithra_raghu · x · 2026-11-21

Samaya AI ran a live Q3 earnings-season experiment with Fortune Magazine, predicting results for 50 companies across 10 sectors. It outperformed consensus on 40/50 predictions and called beat-or-miss direction correctly 45/50 times, using only public information. Examples: it predicted Verizon's lower equipment revenue from longer phone upgrade cycles (43 months), caught falling lumber prices in Weyerhaeuser's 8-K, priced in Vertex's $10B acquisition costs, deviated from Palantir's guidance based on Q2's 18% growth, and used realized oil prices for Permian Resources. Founder Maithra Raghu frames this as early evidence of an AI-driven investing wave, with firm-specific judgment agents as the next step.

Related event: Samaya AI Beats Wall Street Consensus in Earnings Test(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →