Chris Patil

@donotgogently.bsky.social

Biologist, science communicator, AI dilettante, foodie, OG em dash enjoyer

Bit of a tangent, but it's funny how in so much of sci-fi, FTL (which to the best of our knowledge is literally impossible and should break causality) is seen as a matter of course, while stuff like curing aging (which is merely an incredibly hard, complex problem) is a bridge too far.

I have grown a pet peeve against the term "AI slop." It is such a lazy term. It is deliberately not attributing something being "slop" to the wrong entity. It is _entirely_ possible to use AI to make great things. The problem isn't AI. The problem is human laziness.

I was today years old when I learned that in prime-agent you can navigate from your current session to any of your other sessions and child sessions using the arrow keys ⬇️⬆️➡️⬅️

always striking when folks who are famous for being imaginative are like “yeah I’m not going to be imaginative about this specific topic or learn new information in order to better understand it”

Bild

does anyone know how a model gets evaluated by Artificial Analysis? Does AA just pick important models, or does the model developer/sponsor pay AA for the cost of tokens, or [some secret third thing] (I am interested in seeing how Mercury 3 stacks up on the benchmarks)

Inspired by a question from my son, I had Claude put together something I've always wondered - how long does it take for a supercomputer to make it to your home?

A graph with floating-point operations per second on the y-axis and year introduced on the x-axis. A purple line is for supercomputers, starting with Cray 1, going through Cray 2, the Connection Machine, and so on up to El Capitan. Then there's a blue line with PCs and gaming machines, starting with a 486 PC, going through various PlayStation generations, and then ending in an RTX 5090. There's a gray shaded band that roughly separates those two lines and shows that home machines trail supercomputers by about 17 years.