Mohit Kulkarni

huge. @NeocambrianAI is upto some amazing stuff

Introducing: Cohere Command A+ We’ve created our most powerful LLM yet, optimized it to run on as little hardware as possible, and released it open-source for all.

@pHequals7 crazy

meet Swift's 4th most trending developer 🤪

@kukreja_abhinav any blogs/people i should follow to keep up?

If reading about developments in AI is making you nervous, for the love of God, do not even try to keep up with what they’re cooking in robotics.

This is pretty insane for the model size. GDNs really are super interesting architectures.

@pHequals7 GLM 4.7 is good

what are the best open weighted models that natively support multi turn agentic steps (reasoning, tool calling and structured outputs) anthropic, openai and grok seem to be the most well documented ones but are pricey struggling with deepseek and minimax (on openrouter)

@pHequals7 given enough friction still remains to create software for the next few years, indian IT services could actually benefit by commoditizing cheap labour+cheap tokens. Id suppose this friction to exist for 2-3 years, but extremely hard to predict. GPUmaxxing is needed regardless.

this and daniel gross's agitrades essay keeps ringing in my head "If you’re India, for example, where double-digit percentages of your GDP are literally IT services, what do you do when Claude and GPT-5 tokenize like vast portions of that flow" GOI should be GPU-maxxing

Growing increasingly concerned about AI efforts in india.

@paraschopra Another very cool and useful way to think about dimensionality of data is the Johnson–Lindenstrauss lemma. Basically says that any set of n points can be mapped into k = O(log n) dimensions while almost preserving distances, and hence also angles

Learned something very interesting today! Random projections of a non-linearly separable data onto high dimensional spaces is enough to make it linearly separable. Consider a dataset like XOR that you can't linearly separate. Now, if you project each 2D point onto a D (=50) dimensional space using *randomly* initialised basis vectors, each direction creates a tiny difference between the classes (e.g. gives 51-52% accuracy) because expectation of two classes differs slightly when randomly projected. So each randomly projected feature becomes a tiny discriminator and when you aggregate it over 20-50 such discriminators, a linear classifier is able to separate them perfectly by simply learning how much to weigh each feature. One intriguing possibility of this is that we're able to train deep networks because random projections make most of the data already separable, making the job of gradient descent easy.

@kukreja_abhinav Congrats!

Starting the year off in our new office! ❤️☺️ happy new year everyone!

@Gravito841 Whats 99.99%tile for people aged 20-29. Couldnt find any statistics for india/worldwide

1/6 Introducing Seesaw: a principled batch size scheduling algo. Seesaw achieves theoretically optimal serial run time given a fixed compute budget and also matches the performance of cosine annealing at fixed batch size.

SSMs promised efficient language modeling for long context, but so far seem to underperform compared to Transformers in many settings. Our new work suggests that this is not a problem with SSMs, but with how we are currently using them. Arxiv: arxiv.org/pdf/2510.14826 🧵

@vikhyatk github.com/NvChad/NvChad more opinionated but better defaults

what is the omarchy of neovim configs? is lazyvim good?

New preprint! We study in-context learning (ICL) through the framework of task alignment: how well do pretraining tasks match the test task distribution? arxiv.org/abs/2509.26551

Title: Advice for a young investigator in the first and last days of the Anthropocene Abstract: Within just a few years, it is likely that we will create AI systems that outperform the best humans on all intellectual tasks. This will have implications for your research and career! I will give practical advice, and concrete criteria to consider, when choosing research projects, and making professional decisions, in these last few years before AGI. This is my current go-to academic talk. It's mostly targeted at early career scientists. It gets diverse and strong reactions. Let's try it here. Posting slides with speaker notes... -- The title is a play on a very opinionated and pragmatic book by the nobel prize winner ramon y cajal, who is one of the founders of modern neuroscience. To get you in the right mindset, on the right we have a plot of GDP vs time. That is you, standing precariously on the top of that curve. You are thinking to yourself -- I live in a pretty normal world. Some things are going to change, but the future is going to look mostly like a linear extrapolation of the present. And the plot should suggest that this may not be the right perspective on the future. This plot by the way looks surprisingly similar even if you plot it on a log scale. We didn't stabilize on our current rate of growth until around 1950.

@kalomaze Seems a bit too much to call a scientist old just because you disagree on one thing. Is it that hard to believe that LLMs arent the endgame

sometimes people just get old and that's okay

@Gravito841 Isnt this ur batch?

@Gravito841 Damn i miss iitk its like the funniest place ever

@Gravito841 This is like darwinian natural selection for hall 2 kids. You cant take that away from them