Shubh Srivastava

I've taken part in a hackathon. Cheap laptop. 6 months. The hardest AI problem. Previous year winners are researchers with decades of experience. I'm just a graduate. I will be a case study on how far one can go just on sheer will. I don't care if I lose. I won't stop.

okay chatgpt is fucked up. We are back to emojis. They really did switch back to gpt-4 I'm sure. AGI they said huh?

@jyobo10 damn. Need to use this asap haha.

learned you can add a code snippet to your github profile couldnt think of anything better than this

@ThedatagGuy numpy. Be good with thoery part. I started it 5 days back tho and have solved 17 problems only so I don't think I can give any advice lol. Maybe we can discuss after 2 months.

@TheSuperEng Pytorch or numpy ? Which one r u using How are you tackling the problems I am also doing it from the last few weeks

Going to neutralize these problems today and finish dataset imputation techniques.

I've finally started to learn more through documentation than through yt videos. The more you program seriously and the more complicated the programming concept is, the lesser the videos on youtube are.

@som23x @drummatick It will and it should. Different countries should also be involved.

@TheSuperEng @drummatick When DeepSeekv3 launched, the West strategically moved to restrict NVDA chip exports from the US as it became very powerful. If similar progress is now happening through Chinese alternatives for Mythos, just imagine the impact 😂

"DeepSeek is a threat." "It's a security concern." "It's Chinese." Dude, go ahead and ask American companies to do more open source. We have cheaper models, free access to open weights, and don't need to rely fully on monopolies, thanks to Open weight models.

@encodecoder deep-ml. leetcode style problems

@TheSuperEng What platform is that?

I spent 4-5hrs on this problem. I could've seen the solution in under 2 secs. But, I didn't. Major problem I faced was different notations of linear regression equations. I had studied through algebra but, the solution to this was written in matric notations which didn't go in my head easily.

@LexnLin @TheSuperEng can’t wait to play it!

@vidit_odedra yeah krish naik's courses are based on ng only. pretty good imo.

@TheSuperEng ahhh, makes sense; i learned all of the Fundamentals from andrew NG. would not recommend going through the entire course. yr approach is better. but the content he covered was really good.

Today big day lads. Gotta grind dataset missing values fixing. - KNN imputer - linear imputer and more.

Yes I started with golang. But, I didn't want to restrict myself to go and backend. So then explored frontend (react, typescript etc.) then built something with it. But, now I want to build even more difficult project which is based on MLOps so have to learn ML for that. Have to make time for all of this separately mate.

@TheSuperEng You were learning go back then right so how you manage this hops from one stack or domain to other mate. Also are you working somewhere ?

Worked on these three problems today. These problems are for iit/stanford kids tbh. Way too iq demanding imo. Some easy problems are hard asf while some are easy. Trying to solve 3-4 problems daily but, will take time until I pick up speed.

@vidit_odedra not single course.

@TheSuperEng where are you studying all this from ? a single course or randomly ?

@distroaryan You gotta do whatever it takes you know. Even if it primitive method.

@TheSuperEng Same bro, paper par likhne se dimaag Kaam karta hai

Don’t ask me but I’m programming on paper now lol.

@distroaryan It's really interesting field. I'll study that later too. Give it a try.

@TheSuperEng I have also explored a little about deep learning but never went too deep into maths 😅

@drummatick Much of recent startups are agents.md file in summation. Even yc is bullish on this. Ain’t no way this working out.

My prediction is, slowly everyone will realize that using SKILLS to implement a predictable workflow like browsing, checking logs, or testing use just wasting useless tokens and just having the raw code is better than have SKILLS.md later converted into the code

@abdullah_twt23 deep-ml website

@TheSuperEng Is this on kaggle ? Sorry I don't know