KAT-Dev-32B 32B-parameter model for software engineering tasks On SWE-Bench Verified, KAT-Dev-32B achieves comparable performance with 62.4% resolved and ranks 5th among all open-source models with different scales vibe coded a quick chat app in anycoder
RESEARCH
-
Pre-training and Fine-tuning Share Same Function in LLM Book
By
–
Fun fact: In my Build A Large Language Model From Scratch book, I reused the pre-training function for the supervised instruction fine-tuning chapter to show this as clearly/intuitively as possible. Only the dataset (structure) changes.
-

Professor Newcombe’s 1903 Critique of Aeroplanes: Power and Safety
By
–
Quotes from Professor Newcombe from 1903, where he criticised aeroplanes: 1. Animal muscle vs. machines: the power source is fundamentally different. We have to substitute “crude and clumsy” engines and batteries 2. The aeroplane’s safety problem: sustained only by speed, with
-
Training Data Determines Model Behavior and Outputs
By
–
More generally, it’s whatever is predominant in the training data.
-

Memory Imagination and Human Perception in AI Context
By
–
“Most of us are a living, breathing soup of memory and imagination – and we may not be the best arbiters of which is which.” Lines like this are the reason I love reading Arundhati Roy!
-
Creating Concise Languages That Map to Popular Targets
By
–
The best you can do for now is create a less verbose language that maps onto a widely used target language. You can then generate paired training examples programmatically so the LLM learns to operate with fewer tokens. (Congratulations, you also just reinvented Triton.)
-
LLMs Benefit from Human-Designed Programming Languages
By
–
LLMs are quite capable coders. Now, people say they are limited by programming languages designed for humans. I think it is the opposite: they are enabled by programming languages designed for humans. You cannot just design a new programming language for LLMs because where
-
Share AI, Mathematics and Machine Learning Insights
By
–
If you found it insightful, reshare with your network.
— Akshay 🚀 (@akshay_pachaar) 27 septembre 2025
Find me → @akshay_pachaar ✔️
For more insights and tutorials on AI, Mathematics and Machine Learning! https://t.co/DrcSzcaOH4If you found it insightful, reshare with your network. Find me → @akshay_pachaar For more insights and tutorials on AI, Mathematics and Machine Learning!
-

India AI Impact Summit 2026 launches YUVAi Global Youth Challenge
By
–
India is hosting the #IndiaAIImpactSummit2026 on February 19–20, 2026, in New Delhi. ‘YUVAi: Global Youth Challenge’ is one of the flagship side events of the Summit, for which applications will be opening soon. The program will offer shortlisted students a 10-day technical
-
Data, Evaluations, and Ablations Require Substantial Effort
By
–
data, evals and ablations take a ton of effort
