Bhardwaj Rish

The more I use LLMs, the more I think they're just like us. We try to differentiate them from us, but I find far more similarities than differences.  "AI is merely fancy autocomplete" Oh yeah? As a human, are you any different? When you start a sentence, before you can complete the sentence, can you tell me what the last word would be? Nope! Do try though, it's a fun experiment. "AI is terrible at Math" Yeah. And so are most humans. Just in the same way I can't ask you what 56 x 97 is and expect an answer once you've learnt how to speak a language. As a human, what do you do to learn Math? You first learn the fundamentals. You learn through repetition the same concepts over and over because they're hard. You don't talk  math, you try to feed into your brain  how math works. For mental arithmetic, you practice over and over. You memorize the tables by heart. You have a separate training to get that autocomplete right. "AI is confidently w

Consensus building takes a lot of time. Decisions take exponentially longer to arrive at with each additional person involved. And because big tech works in a fashion where everyone has a say and each say matters , decisions take ridiculously long. And this is where I've concluded: Big Tech needs leaders. You need a leader sitting at the helm of each product who dictates  the direction of the product. The whole org working under them just worries about how to make it happen , rather than aligning themselves on what decision to make . I highly respect leaders who: Listen to all arguments. Think from first principles, so that their decisions hold value when tested. Make decisions that are self consistent and based in logic. Make opinion based decisions unwaveringly. Decisions often come down to making a tradeoff. Have the magic sauce. That is, they generally have a good sense of intuition. Are transparent about their values. Say: We're doing xyz . Get it done. Naturally, the le

Nisar works at the Ministry of Road Transport. If you don’t know much about government jobs, there’s surely one thing you know - things move at a snail’s pace. I met Nisar after many years, and naturally we got to talking about his job. “I’ve been in this department for 7 years now. 7 years! It’s incredible how much power you have once you’re in the system”. He speaks of his job very passionately. He’s the lead ML Engineer in the Ministry, having created the position when he first joined. He has since led initiatives for where roads should be constructed based on all the data the government has collected about public movement. He describes how his team decides where the roads should be built.  He tells me how they end up using really simple ML models, and most of the things can just be done using regular Excel Sheets (!!), and that the core engineering isn’t actually very difficult.  “Doesn’t that bother you though? That you you're far from doing sophisticated engineering, and th

You’re a team member of a construction company and you’re tasked with solving a discomfort for commuters at a three-way intersection. It’s a moderately busy street with pedestrians, cyclists, cars, and trucks all crossing by, and the way the traffic lights are configured result in everyone waiting far longer than they ideally should. Now the situation is in no immediate need of fixing - it’s a mild discomfort, but the people accept it for what it is. However, the city is growing and your team knows this intersection might be a cause for concern in the future. (Side Note: Mini Motorways is just such an aesthetically pleasing game!) What do you do? Well, there are two approaches to solve the problem at hand. You modify the current traffic lights configuration with the caveat that the road will be in need of change once the city grows. You make an extensive, elaborate plan to understand what the future potentially holds, and preemptively fix the situation before it ever possibly arises.

There's a Bollywood movie in which the protagonist studies the life of Mahatma Gandhi for weeks on end spending night after night in the library - till one day he starts talking with Mahatma Gandhi himself - well really just a hallucinated version of him. And through the movie, the hallucinated character helps the protagonist lead a life of Gandhigiri. He also helps him answer peculiar questions of Gandhi's life - which one would agree is a cool thing to have around. Except, one day the antagonist asks the protagonist to ask Gandhi a very specific question - a thing only he would know.  At that point, his hallucination goes silent. To me, that's ChatGPT. It's really helpful with things I already know about. It can write scripts which would otherwise take me an hour to write. It's also great for normal chit-chat and asking it to be an unbiased judge of morals and values. But for things that are very specific, like how to get Bazel to include the exported targets in t

"I feel tired" I said. "Then stop feeling", he said.  This is one of my fondest memories of my first job. My mentor just casually slipped it in, but it's a constant reminder that I have a choice . - Cold showers. The Internet says it's healthy for your skin. Huberman adds a bunch of biology and says it gives you a 4 hour dopamine effect. But more than anything, I say it's your mind's win over your body. It's been half a year now, but every day I step in to the shower, my body doesn't want the cold water. It sends signals to my brain saying this too much, this is too cold. I start with my feet, go up to my hips. And then my body begs me not to pour any over my chest. My innermost fears of visualizing a shark in deep waters surface the moment I close my eyes. Every single day. And yet when I pour the cold water all over me, I feel great. I realize I can just   decide  the water isn't cold. And then it isn't. I don't shiver, I truly

If you're like me, a programmer taught how to think in terms of relational databases, you just can't design a NoSQL. Wait. No one can design a proper NoSQL database for relational data because it's a data store built on tradeoffs.  I have very relational data that I am being forced to feed into Firebase just because of how convenient Firebase is to set up. Here's a simplified problem statement: There are User s, each of whom has an email , name , and gender . These users can be friends , and have to consent  to their friendship - sent , pending , and accepted . Create a NoSQL json table such that there is no duplication of data, and each user is able to query a list of their friends including invites with all the user details. Your tradeoff is effectively choosing between optimizing one of either read or write . No matter how you structure your code, you'll end up with either: Optimize reads, data duplication ( View Code ) Each user contains all of their friend&#

“You know we’re still friends because life has it this way” she remarked, “there are certain things in life you chose. That you have complete control over and no matter which way the river flows, you’re always headed north. Then, there are other things that just work out in your favor. Like our friendship. You live in Europe and I live in UK, and what’s more when we go back home to India we live half an hour away from each other. How big a coincidence is that. Sure we plan trips and we make time to speak with each other. But that’s all hinged on being in close proximity. The further away we’d stay, the more our lives would change, and slowly our boats would drift apart. Imagine if I had taken up the offer to study at CMU. I’d be a poor student looking for a job. Right now, we bond over having moved away from home and setting up our own lives. As a student you just wouldn’t get my everyday problems. Heck, current me wouldn’t identify with student me. I think that’s it. We’re still close

At dinner in New York, A told me how he waited many years for the right one. He always wanted children. Now he sacrifices his career to pick them up and drop them off to school everyday at age 50. At lunch in the office, B told me how his wife decided to take a step back from her career and take on a part time job. Her mother was never there for her. She wanted to be there for her kids. On our monthly catch-up, C tells me they’re pregnant. He was very happy. He later told me she’s 5 years older than him and that it was a real challenge to conceive. He’s ecstatic. At his desk, D told me he broke up with his girlfriend of 6 years. He’s 32, but he’s moving to California to start a new life. He can always come back to Germany, he says. On a video conference, E told me he’s quitting his job to move to Barcelona. This was the first time in 2 years he mentioned he had a daughter. This was also the first time he said he’s a divorcee. Right before our half marathon, F tells me they’re enga