Aru_sharma

This was prolly the best session that I attended today at impact ai summit @OfficialINDIAai . Just look at the panel @sarahookr @aliceoh @rao2z @raskarmit @SuryaGanguli

Time to go from 1 man team to no man team. Best of luck 🫠

Maths and Physics is where all the AI architectures will converge in few years so buckle up and get better at maths and physics if you want to work on the frontier

@paulabartabajo_ I agree this is somewhat related to the bitter lesson essay

Advice for AI engineers 💡 Stop building LangGraph pipelines. Modern LLMs are so good at tool calling that your hand-crafted orchestration is actually hurting performance. Let the model decide.

Why dafak is @X chat so weird, why I don't get any notifs for people who have texted me back in March or April. @nikitabier please do something

@Laz4rz bruhhh

Weekend reading list as if the weekend was a week long

@willccbb peter doesn't use any compute it's peter's clone that using it

peter prob uses way less compute than the average oai researcher

@getpeid byeinterview.com

name a tech product that has real taste not expensive, not minimal, not pretty actual taste

If you get caught submitting AI slop to arxiv, the punishment should be generational aura loss. "Dad, why do the other kids say I have no rizz?" "Well son, let me tell you about a mistake your grandfather made decades ago with a model called Codex...."

@justinskycak Very much needed I was struggling with stats lately will follow these tips fs

1. Do the problems. Consuming information is not enough. 2. Treat confusion as diagnostic information. 3. Master the prerequisites. 4. Automaticity unlocks deeper understanding. 5. Treat math like training, not cramming. 6. Space out exercises of the same kind. 7. Most "motivation problems" are actually sequencing problems. 8. Review like you mean it. 9. The broader goal is to get good at something hard.

@anonymashed_ But we can review that code in open

@Aru__09 But it’s not a reliable signal, people fill it with coding agents

Resumes are ChatGPT. Cover letters are Claude. Portfolios are Cursor. Every hiring signal is noise now. How are you actually evaluating candidates?

The more I try to interact with people, the more I realize that success often isn’t just about grit or hard work it’s also about awareness and exposure. A lot of this comes from the people around you: seniors, friends, peers, and mentors. Tier 1 colleges naturally provide strong peer groups and clubs that are already optimized for this kind of ecosystem. But that doesn’t mean opportunities are limited to those institutions. If you’re not from one of these colleges, you may need to put in some extra effort, but you can make up for it by actively participating in communities, networking at events, and reaching out to people through cold emails or messages. I got into programs like GSoC and SoB partly because I was fortunate to meet the right mentors in college. But before that, I got into LFX completely on my own. My first attempt didn’t work out, and at the time, no one in my college even knew what LFX was. Looking back, it feels rewarding to now see more people from my college applying, and hopefully getting selected in the coming iterations. Open source runs on the spirit of paying it forward, and I want to contribute to that as much as I can. “Future is open,” and sharing what I’ve learned feels like the least I can do for everything open source communities have given me. If you ever want to know about anything related to OSS, feel free to comment below or reach out via DM. I'll try to help you out in whatever way I can.

In-context learning in LLMs

Maturity is when you read casual as causal. period

Looks neat

@CattardSlim They sorted out things in O(1) lol

China, 30 seconds after Trump entered AF1. 😭

If you think you are smart try solving these, you might get humbled projecteuler.net

If you’re overthinking whether your idea makes any sense, just know there is a flight from Mumbai to Navi Mumbai via Delhi

Reading a textbook without doing the problems is like reading a workout plan without doing the exercises.

What's the best way to network in SF? Is it different from how we network in BLR??

Last year I was unsure if I should go for GSOC or SOB but I thought it would be a good challenge to do both so I did. Fast forward to today exactly 1 year from now, the question is similar but this time it's about mentoring the new contribuors at Gsoc and Sob. Last year I had a choice but this time it's my obligation to pay it forward. All hail OSS.