LLM 0.32rc1 release adds content-addressable hash IDs for messages, de-duplication, and support for new GPT-5.6 models.
Analysis testing if AI labs trained models to draw pelicans on bicycles, finding no evidence of bias.
Explains how LLMs learn low, medium, and high-effort reasoning modes, covering training and inference techniques for controlling reasoning effort.
Analysis of GPT 5.6's 72 configurations, mapping model choices and reasoning effort to performance and cost.
DeepSeek Sparse Attention implementation from scratch, featuring a learned sparse pattern replacing fixed attention with a lightweight indexer and selector.
Explains DeepSeek Sparse Attention implementation from scratch, replacing fixed sparse patterns with learned ones using a lightweight indexer.
A guide to transforming pretrained LLMs into text classifiers, with insights from the author's new book on building LLMs from scratch.
Learn how to use Python libraries like Click, Jinja2, and GPT to enhance and build AI projects and applications.
A technical guide on reclaiming unused swap partition space by expanding the root partition to the left using tools like sfdisk and gparted.
A tutorial on building a generative transformer model from scratch in Julia, trained on Shakespeare to create GPT-like text.
Explores using prompt-hoisting to improve GPT-based code generation, turning tests into prompts to reduce AI's indeterminacy and ensure code correctness.
A technical guide to implementing a GPT model from scratch using only 60 lines of NumPy code, including loading pre-trained GPT-2 weights.
A technical overview of the evolution of large-scale pre-trained language models like BERT, GPT, and T5, focusing on contextual embeddings and transfer learning in NLP.
