papercode is up for sale at @trust_mrr trustmrr.com/startup/paperco…

i wanted to find creators who could promote lattedit.xyz, so i used Treg + Jev to find 30 people who recently posted about Overleaf, LaTeX, and AI research writing tools. got their profiles, relevant posts, public business emails, and a ranked list based on how good a fit they were, along with the total search cost, and reached out to them truly one of the best tools i use on a daily basis now :)

cash on delivery

things i wish someone told me before placement season: > solving 500 problems means nothing if you can't explain 50 > recognizing a pattern is very different from implementing it under time pressure > edge cases will humble you more than hard problems > debugging a wrong solution quickly is a skill of its own > revising problems you've already solved is far more useful than chasing new ones > being able to explain your approach clearly matters just as much as getting the AC if your placement season starts next month, pick 30 problems you've already solved and redo them without looking at your previous submissions. time yourself. you'll be surprised how many you thought you knew.

I don't think this is Instincts' fault. This is because LinkedIn hates any kind of automation. Even if you had written JavaScript console code to automatically manage that, you would have still been blocked. I don't really feel this is Instincts' fault.

Thanks, everybody, for the suggestions. I am actually going to log everything in detail. I'm starting to watch all of them. cneuralnets.xyz/animes

hell nah we're losing

RAPHINHAAAAAAAAA

Chelsea is not mid level

the idea is simple. traditional decoding methods like Top-k and Top-p use fixed sampling parameters, regardless of how confident the model is about its next token. Top-b takes a different approach. it uses Shannon entropy to dynamically adjust the range of tokens that can be sampled. when the model is confident, the sampling range becomes narrower. when it's uncertain, the range expands to allow more diversity. they experimented with Gemma 3 4B on GPQA and GSM8K, comparing Top-b against Top-k, Top-p, Min-p, and other sampling strategies. the results were pretty interesting. on GSM8K, Top-b achieved 26.80% accuracy compared to Top-p's 26.13%, while reducing variance across runs by almost 10x. on GPQA, it also achieved the lowest average generation entropy among the tested methods while maintaining competitive accuracy. what i found interesting was how a relatively small change in the decoding strategy could make generations more consistent without retraining the model. pretty fascinating how much of an LLM's behavior depends not just on its weights, but on how we actually sample from its probability distribution.

Hey guys, suggest some niche animes with less than 12 episodes total that you love and that I would love to watch.

day one to one day 😊🙏

Liquidated all my stocks and invested in the only thing which will never disappoint me

Today I booked a Rapido back to college. I booked under AC priority or something (AC or AC priority, I remember) but the person came and he said, "Sorry bhaiya, hamare paas AC nahi hai. Non-AC hai. Non-AC pe booking nahi ja raha hai isliye main AC pe book kar raha tha." I was obviously pissed off, right? It's not about the money. It's more about this: if I'm trying to get AC, I will. I'm ordering AC for a reason, right? Then I got pissed off but to be honest he was very nice. He explained a scenario: his AC went bad and he needs 30,000 but he does not have the money. If he does not have AC he's not getting bookings so that's why he is in a fix. I was like, "Okay, this is a well-behaved person, right? Okay, not a problem, let's go." We went and I gave him almost 1.75 times whatever the fare was. He started crying, "Sir, bhaiya, what was the need of this? You should not have given. I have cheated you." I was like, "This is one of the few people who has a conscience. He was feeling bad about that." I was so elated. I was like, "Please keep the money. Don't need to tell me anything but you would need your AC to be working again, I guess, so keep the money." He was so happy, was almost crying, and was at the same time guilty. I loved that, to be honest: this guy was honest about everything and was nice throughout the whole thing. People might have problems and if you just be nice to other people, nice things will happen to you. That is what I believe.

I was going to order something online, which is maybe 2% of my whole total credit limit. It kept on getting blocked for security reasons. Bro, I am the only one ordering. Nobody even uses my card.

A 1 TB SSD is as pricey as an above-average gold and diamond ring What is this memory economy?