Building a Toy GPT from Scratch in PyTorch
Learn how a GPT-style transformer works by building one module at a time in PyTorch. No prior Transformer knowledge required.
Personal blog of Arun Kant Sharma — essays on programming, web development, and learning by going back to first principles.
Learn how a GPT-style transformer works by building one module at a time in PyTorch. No prior Transformer knowledge required.
A practical, from-zero guide to writing your first CUDA kernels, understanding the GPU execution model, and optimizing with shared memory, tiling, and streams.
Claude was silently adding itself as a co-author on every commit. Here's how I scrubbed it from my entire repo — branches, tags, and all — using git filter-repo.
In-house AI agents are seductive but fragile. Here's the case for using agents to explore and concrete workflows to ship.
Building a modern React app from scratch - no Vite, no Next.js - to demystify the magic of `create-react-app` and its friends.
A dump of articles and links I've collected in my browser, with brief notes - part 5 of 5.
A dump of articles and links I've collected in my browser, with brief notes - part 4 of 5.
A dump of articles and links I've collected in my browser, with brief notes - part 3 of 5.
A dump of articles and links I've collected in my browser, with brief notes - part 2 of 5.
A dump of articles and links I've collected in my browser, with brief notes - part 1 of 5.