Claude 3.7 is the best model for devs.

Bro has a new favourite model every couple days

Waiting for fireship video

Every fing week a new model and a new bar chart gets introduced to the world.

I used Claude 3.7 yesterday without even noticing it. I thought I was interacting with 3.5. I asked it a question about my code, it failed miserably, I had to roll up my sleeves and do it myself. Business as usual.

I was skeptical until I saw they made graphs with the big line that goes bigger than the big line that's not so big but still big, except it used to be that the bigger line now wasn't as big as the bigger line last time which is the line that's less bigger this time if you don't believe me just look at the lines on the bars charts graphs look at the numbers they mean things I promise the numbers will never let you down, trust the numbers, be the numbers, the numbers will make us human, the numbers will make us whole long live thenumbersthenumbersonthegraphonthechartonmyeyelidsasilaymedowntosleepipraythatmyapikeydoesntleak

btw, AIME isn't "high school math" really - yes, it is a competition for high schoolers, but it's the ones that are best at math, the peak of the peak math nerds. i was a math nerd in high school, but the problems in there still are often above my level

They've won for 5.2 days till the next LLM drop

Happy 400K, Theo!

Yes! please do a cursor video on the new release! I got early access a couple days ago and have no clue about the new stuff. And I couldnt find any info online about it. Would also love to see your insight about how u use cursor in general. Like when to command k, when to use composer, etc

I like to think of "non-reasoning" LLMs as "intuition" layer and "reasoning" LLMs as "thinking" layer and use them like I would use my brain when solving problems myself. Most things I don't need to "think" about, relying on the "first thought" or "intuition" is more reliable and only if I see that my intuition was wrong a couple of times, I would start "thinking". Or if I can tell in advance that it's better to "think" first instead of just relying on "intuition".