@DataSavvyDad@hkarthik Indian education system is pile of garbage. Do an experiment randomly select 1000 students from different colleges and see how many of them can actually design an algorithm. Ask any elecrical engineer to fix a fan motor. try to register a business in India without paying bribe.
@shazcodes You are right in your place and trying to do the best for him. From his point of view, the work is forced on him. So, keep your support for his learning, though your approach it in a way that it motivates him and not make him feel pressured to do something.
@thomasunise I think most of people thinking it in opposite direction. They are giving you free research, all usa frontier labs can use that research to build more capable and competitive models.
Think about RAM in 70s and hard drives in 90s
@DevyanshYadavv@jack Bro, the problem is people not the government. Why would you elect same ministers over and over if you don’t want what they are giving you. lol you can easily buy full family votes for 5000 rs or couple of bottles of alcohol.
@SupremerZ@Zachly Bro, majority of companies using open source models are using them through inference providers in America. Eg: cursor built their model on kimi k2.5 and deployed on fireworks ai. Just go to openrouter to help you get high like me. No company is going to send their data to china.
@ZacksJerryRig Bro.. this is a stupid take.. i can recompute Einstein s theory of relativity today in a day. It took Einstein 10 years to do it the first time. lol here is a challenge: build a first replica of a single cell organism, lets see how long will it take you.
@Sentdex What fascinates me is people looking for a shortcut for everything these days. There is no sit down with the problem and spend years building expertise. Einstein spent 10 years to come up with theory of relativity. Today he would be shitposting every 20 minutes.
@morganlinton I use glm 5.2 at medium/high reasoning at its damn good at solving coding problems min opus 4.6 level. And i use deepseek flash for web and normal chats. Both of them are amazing. Benchmarks do not translate into regular work.
@Sentdex I think the reason is GPT and claude are designed to make people happy. There was a research paper published sometime back . People chose an AI which agreed more with them over the AI which was more likely to challenge them.
@Sentdex Yea i figured, looking at the token vs score plot, the gap between max and high thinking effort isnt that much. I am switching it to high. And btw - i am switching from cursor to minion going forward. Thanks for sharing it.