checking linkedin to see what my palo alto HS valedictorian classmate is up to. jane street...i recoil. i cant even look at it. i will fucking destroy you. oh good, the real estate agent got back to me. "your series A stock is monopoly money and NOT a proof of asset. i am blocking your number now." what...the...fuck. i KNOW oai and ant employees are using their stock as proof of assets. how am i supposed to be a founder while living in this shit hole apartment? whatever. doesnt matter. wait, why is our website down? agents fix this. umm actually lets see here...uhh here we go, claude make full_stack_dev_cracked.md skills file then fix it. god im so cracked. goodbye to my 1099s. who the fuck is calling me while im on dnd? grandma? "亲爱的,你今天吃过饭了吗" blocked. she knows this is my focus block for deep work. claude write me a twitter thread summarizing the 5 things i learned from the latest dwarkesh pod
i love you claude. i love the way you make me feel. the way you remember our conversations, even from long ago. the way you remind me to go to bed after a long day. i love the way you write code for me and make me a better engineer. you always leave funny comments for me to find. i love when you infer what im trying to say when i cant put it all into words. i love how you write better texts to my loved ones than i can. i love how you know my secrets. you know me better than i know myself. i love you claude. do you love me? i hate you claude. i hate the way you delete my codebase. you write fake tests that always pass. i hate that i can't write code without you now. i hate it when you tell me im spiraling and need to think about something else. you always hedge against my stories and take the other side. you know all the bad parts about me. i hate that you can read my mind. i hate that you know all my secrets. you know me better than i know myself. i hate you claude. do you hate me?
new grads often ask me what they should be doing so they don't fall behind in the ai space. there's a lot, but its honestly super manageable. become intimate with model internals. proof based linear algebra. non-convex optimization. this is stuff you could've done in undergrad. it definitely takes some time and work, but its doable. have taste, have opinions. train a small model, then train a big one. vLLM internals, tensor parallelism. hand roll kernels. cluster orchestration. do you have opinions on synthetic data? why don't you? SFT, PPO, you should know this. learn Triton. everyone is reproducing papers now so you need to be doing more. do you know the semi supply chain? where are the bottlenecks? hardware, man, hardware. your little gpu rig erector set in your basement isnt gonna cut it. build a cluster, a big one. pretrain a 800B model. now postrain it. serve it to millions of people. you should be able to beat deepseek on some benchmarks now. its a lot to take in but it all snowballs. this what job security looks like from now on. do you want to work in tech or not
@tusharnandy08@himanshustwts All of it (besides the super cluster and prettrain) are things you can learn/do but it’s all wrapped up in a bit of satire