Six months after publishing 'Large Language Models for Mortals', the author recompiled the book's 264 executable Python code snippets and found mostly stable APIs with only three retired models requiring updates (Claude Opus 4, Claude 3 Haiku, Gemini 3 Pro). Core API components like chat completions and RAG patterns remain relevant, though AWS Bedrock setup has simplified and open-source models have improved significantly, creating potential gaps in the book's coverage.
A developer shares progress on building a vintage language model trained exclusively on pre-1900 texts. After three and a half months of intensive work, they developed new datasets, three new models, a benchmark, and an evaluation pipeline, improving upon their initial 340M-parameter base model that could generate Victorian prose but couldn't maintain conversations.
A researcher at AI2 describes their transition from quantization research to coding agents, detailing how a small team of five researchers and 32 GPUs developed Sera, a method to finetune large language models on private codebases for efficient coding agent deployment. The work eventually scaled to 96 GPUs and enables cheap specialization of models rivaling larger teacher models on private data.