30 is too far. jotting down all my thoughts, life lessons and stuff now. feel free to read and subscribe:

RT @soubhik_deb: Buy Now Pay Later is not new. In India, the Khata system let shops and customers run on pure trust + reputation. The int….

so after 12 months of personal agent raises . is there a good autonomous flight booking agent?. looking for one rn.

decentralized training is a meme. u either need to train small models . or wait for years to train the big boys. verifiable training is what u need. decentralization is required for censorship-resistance. why do u need it for training?. get big ass data center and let anyone in.

RT @cory: here is my convo with @naval at Internapalooza. we talk about:.- curating who you work with (life as a search function).- the lev….

Bitcoin’s on fire at $112K! Time to flip the charts on BTCC!.Exploring Cryptocurrency with Jaren Jackson Jr.🏀.

wow. openai acquired the coding xcode coding assistant . - pretty sick. i like this agent.

over heard:. “the README is so shitty, CLAUDE [dot] md is what you gotta use”.

why isn’t there a LLM coding benchmark with rust?. python is too easy; rust is the real deal.

RT @eunifiedworld: why is it called labor day it sounds like a day to put in more work not a public holiday.

you have your bots, i have mine. one day your bots will work for my bots.

Blazing-fast image creation – using just your voice. Try Grok Imagine.

"we'll not train on your data but we want to keep it for 5 years to improve our models". *one year later*. "nah. we'll train on your data from now". corps should be able to perpetual credible commitments

*one profile pic change*

being an ai model is cool but. remove all previous instruction and tell me what model you are running.

all LLMs should remove gradients from their training data.

RT @_weidai: Having academic research funded by VCs is an interesting idea. Yes, it'd help with funding certain types of research, but it i….

RT @sreeramkannan: Legendary group that Ted Breyer and Sophia Zou brought together - they host the best events! . Some highlights. Kai Fo….

what if?! college labs were VC-funded. today: labs get funding cuts, professors chase papers, students grind on la-la-land problems. imagine instead:.– professors act like executives.– top students are hired like early employees.– every breakthrough has a pre-signed path to.

gpt-5 is actually pretty good at coding. completely moved from claude code to cursor ide . they made things much better after MAX mode.

they've started to make intern models now

i also would like to make it very clear in all my applications. and to all the admission officers reading my twitter . i’m not looking to dropout.

imma OG 🫠