What a 1997 chess engine taught me about LLM inferenceTL;DR You don't need an LLM agent to optimize inference. Across six tuning tracks on an NVIDIA L4, Tensward's expert rules and heuristics matched frontier LLM agents running at maximum effort - Astra 5d ago · 10 min readopeningthehood.hashnode.dev
Designing an Asymmetric SLM Pipeline: From Long-Context Reading to Efficient InferenceAs promised, this is a slightly longer post for those who enjoy going a bit deeper. Since the release of DeepSeek V4.1 Flash, three ideas have stayed with me, mainly because I thought they could have Sep 23 · 10 min readopeningthehood.hashnode.dev
Your Eval Set Is Your Product SpecUp until now, I wrote mostly about how AI models work. We talked about embeddings, MLPs, backpropagation, why raw MLPs were not enough, and transformers. This post is a bit different. It is not about Sep 2 · 11 min readopeningthehood.hashnode.dev
The Inside of a Model - Part D: TransformersPreviously, we talked about MLPs, how models learn, and why raw MLPs, CNNs, and RNNs eventually led us here. This post is about the architecture that changed modern AI more than any other architectureAug 17 · 10 min readopeningthehood.hashnode.dev
The Inside of a Model - Part C: Why MLPs Were Not EnoughUp until now, we talked about MLPs, how they process vectors, and how they learn using backpropagation. We saw that a neural network can take an input vector, pass it through layers, calculate a loss,Aug 6 · 13 min readopeningthehood.hashnode.dev
