https://meaningness.substack.com/p/probability-theory-understood-meta-rationally i don’t really agree with this criticism of probability theory. there is a coherent way to interpret “probability theory does TOO apply to every situation”. (and it doesn’t imply that humans should go around making numerical probability estimates all the time.)
As the article says, the important thing is knowing what purpose is the estimate supposed to serve. Mathematically speaking, do we want to know the mean, the median, the worst case, the 90% or 99% percentile?
Unfortunately, in my experience it usually does not help to ask your manager about the purpose, because the answer is often “I simply need you to make an estimate”. Then, if you make a 99% percentile estimate, the manager will complain that it is too much; and if you make a mean estimate, the manager will complain whenever the task takes longer than estimated.
links 7/2/26: https://roamresearch.com/#/app/srcpublic/page/07-02-2026
https://en.wikipedia.org/wiki/Single-cell_protein
https://en.wikipedia.org/wiki/Cellulosic_sugars
https://allfed.info/images/pdfs/rapid_repurposing.pdf
https://pmc.ncbi.nlm.nih.gov/articles/PMC9358032/ methane single cell protein as a food source seems to be at the R&D phase
https://www.sciencedirect.com/science/article/pii/S0043135424007218
https://evalevalai.com/research/2026/04/29/eval-costs-bottleneck/ AI eval costs are rising; duplication and poor reliability are problems
https://evalevalai.com/2026/06/30/saturation-blog/ what features are associated with AI benchmarks that saturate faster?
https://lct-app-515466416372.us-central1.run.app/ Live Conversational Threads, a conversation-tracking app
https://drmaciver.substack.com/p/task-estimation-101 David MacIver post on time estimation for tasks
https://meaningness.substack.com/p/probability-theory-understood-meta-rationally i don’t really agree with this criticism of probability theory. there is a coherent way to interpret “probability theory does TOO apply to every situation”. (and it doesn’t imply that humans should go around making numerical probability estimates all the time.)
As the article says, the important thing is knowing what purpose is the estimate supposed to serve. Mathematically speaking, do we want to know the mean, the median, the worst case, the 90% or 99% percentile?
Unfortunately, in my experience it usually does not help to ask your manager about the purpose, because the answer is often “I simply need you to make an estimate”. Then, if you make a 99% percentile estimate, the manager will complain that it is too much; and if you make a mean estimate, the manager will complain whenever the task takes longer than estimated.