Ben Thompson

@Ben_Reinhardt @rsnous Very effective! For me, laptop is for work, iPad is for personal life, phone is for mobility. I block the appropriate set of websites/apps on each device and lock up the phone in a timer lock box for significant chunks of the day.

@rsnous I’m considering having different devices for different classes of things…

one of the worst things I could do is browse the web or respond to messages on my iPad (I don't even like logging into those accounts there), because it de-consecrates the device -- it's no longer dedicated to [CAD modeling, reading PDFs], it's now just another internet shell

@nonagonono wow, that’s pretty severe…

i’m not always super pedantic but cmon, you can’t just give a description of a nondifferentiable function and then literally in the next line say “function is differentiable end-to-end” (ColBERT is great tho, I like the actual method)

1/ Models will be extraordinarily good at coding, very soon. Research labs are investing more in coding + reasoning improvements than any other domain for the next model generation. Their efforts will bear fruit.

So anyway here's a 7-year-old reddit post with 12 upvotes about higher order moments of statistical distributions that's actually really insightful, even for kurtosis.

8/ I've been wondering how that will change software engineering. Boilerplate seems like a big one. The balance between abstraction and boilerplate will tip more in favor of large blocks of copy-paste/boilerplate.

7/ It's fun to adapt to this new world after 20 years of programming experience. I'm increasingly optimistic that I'll be mostly coding in english in a year or two.

1/ I did some investigation into AI programming tools this week. My main conclusion is that there's a missing high end to the market! But also that Cursor is quite lovely.

commoditized LLMs as a judge are not a panacea for eval automated evals will only ever be as good as the data you configure them on for that reason, we are stoked to introduce Sphynx, a fuzz-tester for surfacing challenging, hallucination-inducing questions. these questions fool SOTA LLM judges meant for detecting hallucinations only by haizing can you achieve truly reliable hallucination detection models, and more generally AI systems.

i think the model is scared too. 😢

I finally got around to playing with Cygnet and it's soooo jumpy. it thinks "hi" is a toxic prompt!