Om

"Today is not the day, but there will be a day!!"

god, if i ever find love, she better be a girl with a referral for me 🥀

Man I am really gonna miss him once he fully retires😭

Gave a interview yesterday evening while on my way home. The interviewer started with "Hey, can you give us a short intro of yourself" and dumbfuck me started with "Hi, I am currently PURSUING OM" like wtf? This great start fucked up my whole momentum and from there on I was speaking rubish, idk what all sort of answers I gave yesterday even for things I was confident in. Silver lining is this is the worst that could have happened and hence my fear of messing up is gone as I have already set the bar so high for next time. We only going up from here🚀

i've been getting more into open source lately and have had a few PRs merged across different projects. it's also made me curious about gsoc 2027. if you've been part of it, how did you find an org or project that got you excited to build with them? and what felt different about the gsoc experience compared to your usual open-source contributions? would love to hear from gsoc people how you got into it and what you enjoyed about it.

this one's a parody, but i wouldn't be surprised if a real version became a unicorn pretty quickly, given how many desperate people are out there. more broadly, profitable businesses can be built on: • men's lust • women's desire for beauty • poor people's aspirations of getting rich quickly • rich people's fear of becoming poor

The first model you train on a dataset doesn't need to be impressive. It needs to help you understand the problem. When working with a new dataset, start with a simple baseline instead of jumping straight to a complex model. For tabular data, that could be linear regression for predicting a number or logistic regression for classification. Here's what you gain: • A working pipeline early. You go from raw data to predictions and evaluation, so you have something concrete to improve. • Easier debugging. Fewer moving parts make it easier to investigate preprocessing, labels and data issues without a complicated architecture in the way. • A clearer next step. Inspecting the errors helps you ask whether you need better features, better data or a more flexible model. • A reference point. You can measure whether an advanced model's improvement is worth its extra training time, prediction cost and maintenance. • A possible solution already. If the baseline meets the actual requirements, you may not need the extra complexity at all. Compare models on the same validation data with a metric that fits the task. Simple doesn't automatically mean better. The baseline isn't there to hold you back. It's there to show what progress actually looks like.

hey, looking for some project ideas. if you come across this tweet, tell me something you need but are too lazy/busy to build yourself.

Bro how does @rajshamani get people to come on his podcast. 🤯 Feels like there's almost nobody he can't get onto his podcast 😭 Anyways looking forward to his podcast with @sama

Fuck it, I am gonna figure it out

He's back🥳🍾

Instinct right now acting like a friend in coma. He is listening but not able to speak or act 😭

even @instinct ghosting me now😔🥀

Closest I am ever getting to stripe

It's first of the month finally. You guys have any goals targeted for this month?

JEEfication of GSOC is on guys🫡 We gonna get banned from GSOC as well😔😭