predict-addict.bsky.social
Experienced Data Science Leader | PhD in Machine Learning | 7x Author | Black Belt 🥋 in Time Series | Chief Conformal Prediction Promoter| Mathematician | valeriy.ai
👤 Joined on 26 December 2024
Followers
Following
Total Posts
@predict-addict.bsky.socialjjnevis.com
45xiatai.bsky.social
methiaff.bsky.social
Not Found
slowxtal.bsky.social
catofkyiv.bsky.social
advanced-eschatonics.com
jlin404.com
alxndrmlk.bsky.social
Not Found
Not Found
placidsound.bsky.social
Not Found
chrisecon.bsky.social
digthatdata.bsky.social
mathillustrated.bsky.social
sebastianmaki.bsky.social
knutjaegersberg.bsky.social
isvetunkov.bsky.social
spoont51.bsky.social
geoffnegiar.bsky.social
juanrive.bsky.social
hardymediation.bsky.social
brucemarino.bsky.social
Metrics on retail positioning show the same trap: crowd positioning stays extreme for hours, yet price leads the crowd three-to-one. The signal flags volatility, not direction.
Show PostThis shows up even at small scale. Running real agents in production: 6 agents sharing a task queue generates emergent behavior nobody programmed — retry storms, task starvation, cascade failures. Incentive design problems start at 2 agents. The million-agent version is just this, louder.
Show PostNo, finite-sample exact, not asymptotic as long as exchangeability holds. Intervals may be conservative with small n though. And you can wrap conformal prediction around a Bayesian model to get this guarantee.
Show Post