bakhus.space https://bakhus.space/ Notes on machine learning, language models, and the practical side of using them. en Sun, 01 Nov 2026 09:00:00 +0000 What I check before trusting a benchmark https://bakhus.space/posts/trusting-benchmarks/ https://bakhus.space/posts/trusting-benchmarks/ Sun, 01 Nov 2026 09:00:00 +0000 A leaderboard number is a starting point. A few questions decide whether it means anything for your task. Prompts are interfaces https://bakhus.space/posts/prompts-are-interfaces/ https://bakhus.space/posts/prompts-are-interfaces/ Sun, 04 Oct 2026 09:00:00 +0000 Treating a prompt like an API contract makes it easier to change without breaking things. Retrieval first, fine-tuning later https://bakhus.space/posts/retrieval-first/ https://bakhus.space/posts/retrieval-first/ Sun, 23 Aug 2026 09:00:00 +0000 When a model needs to know your documents, the simplest approach is usually to look them up. A longer context window is not a memory https://bakhus.space/posts/context-is-not-memory/ https://bakhus.space/posts/context-is-not-memory/ Sun, 12 Jul 2026 09:00:00 +0000 Fitting more text into a prompt helps, but it does not mean the model will use all of it well. Evaluating outputs without fooling yourself https://bakhus.space/posts/evaluating-outputs/ https://bakhus.space/posts/evaluating-outputs/ Sat, 30 May 2026 09:00:00 +0000 Reading ten good answers in a row proves less than it feels like it does. Tokens are not words https://bakhus.space/posts/tokens-are-not-words/ https://bakhus.space/posts/tokens-are-not-words/ Sun, 19 Apr 2026 09:00:00 +0000 Why a model can write a sonnet but stumble when asked to count letters. Small models deserve a second look https://bakhus.space/posts/small-models/ https://bakhus.space/posts/small-models/ Sun, 08 Mar 2026 09:00:00 +0000 Bigger is not always the right default. A few cases where a small model is the better tool.