Anup Jadhav

I’ve been looking into running models on constrained hardware and wanted to get clearer on the difference between GGUF and ONNX.

I have been coding with Claude Code for over a year now, and somewhere along the way I have developed a fairly repeatable workflow for getting the most out of a coding session.

AI engineers often need to decide whether a task should use one agent or several.

If you spend any time watching this space, you have watched this vocabulary turn over three times in about a year.

I use coding agents every day, and most days they earn their place.

For thirty years, software organisations were built around the same logic: hire specialists, group them into functions, put managers above them, and build a pyramid on a wide junior base.

In Part 1, I explored the idea that context is not memory.

[TIL post about inference optimisation techniques] A good way to understand speculative decoding is this: Speculative decoding is “draft first, verify second.”

[As part of my TIL series, building an intuition about Q, K, V] A good way to understand QKV is this: Attention is a soft lookup operation.

Part 1 argued that context is not memory.

This post separates three ideas that often get blurred together: the current prompt, retrieval, and durable memory.

Zak Knill wrote a sharp post this week arguing that LLMs are exposing a gap in our standard cloud-native architecture.

TIL from a paper on ads in AI chatbots that putting adverts inside an AI assistant is not the same as putting ads next to search results.

Even though your coding agents are getting faster (and more capable), our ability to evaluate what they produce hasn't kept up.

Part 1 in this series covered capacity planning.