US LAYOFFS ARE RUNNING AT CRISIS PACE: COMPANIES ANNOUNCED 153,074 JOB CUTS IN OCTOBER, NEARLY TRIPLING FROM 2024. IT WAS THE WORST OCTOBER FOR LAYOFFS IN 22 YEARS.

This is easily one of the greatest days on this platform. Seeing all of these MAGA accounts get exposed as foreign actors trying to destroy the United States is a complete vindication of Democrats, like myself and many on here, who have been warning about this.

U.S. LAYOFF NOTICES ARE SURGING AT A RECESSION PACE

M. Bolton Roger Borràs It’s interesting how machine learning advocates think that in typically noisy data machine learning algorithms will discover more signal than more structured statistical models can. Understanding the underlying information content is key.

While I’m not a fan of DAGs in general, they can be helpful at times to simply visibly map out pathways; focus on the indigo pathway here, with correlated other unmeasured genes. This is the challenge with high-dimensional genomics work which seeks to “name names” like AREG

‼️ The computational strength of AI is irrelevant to finding previously unknown information, you just get to the bias floor faster, that’s all. The Cramer-Rao lower bound sets limit for unbiased estimator like RCT, but with “real world” data, AI is basically useless

‼️ 🫧 The asymptotic utility / benchmark performance of greater FLOPS is clear and fits a logistic model quite well. How many LLMs do we really need? Do we really need “AI enhanced” toasters and refrigerators and TVs? $NVDA 🫧 #SaturationKinetics

Jordan Gauthier M. Bolton It is not a good idea to test differences in AUROC because of power loss and degenerate distribution under H0. Demonstrate excellent calibration then use a sensitive measure of predictive discrimination such as fharrell.com/post/impactpo

Ioannis Sagonas MD 🇮🇱 Uriah Finkel M. Bolton The biggest mistake researchers make with thresholds is not realizing that every patient must have a different threshold. Many researchers try to establish all-time single thresholds in the original publication. That’s letting the analyst make decisions for future patients.

‼️ NEW: Trump’s tariffs broke the Economy. Americans (including GOP voters) are feeling the pain and are now telling pollsters all about it: including Fox News. This is brutal for Trump and Republicans.

Sign of a strong economy

M. Bolton Jordan Gauthier Frank Harrell Discrimination = relative (higher risks for events?) Calibration = absolute (are risks 'accurate'?) Dividing all risks by 10 changes calibration but not discrimination. Cf arxiv.org/abs/2412.10288 (accepted) But calibration is more complex, see arxiv.org/abs/2506.17141 (submitted)

Nvidia says it’s not Enron. I actually agree with them, they are Envidia