Nov 09, 2026 The Vision Transformer reads images as sequences of patches Oct 19, 2026 Beyond attention with Mamba, state space models and hybrids Sep 30, 2026 Long context with RoPE scaling, YaRN and FlashAttention Sep 10, 2026 The KV cache budget and DeepSeek's latent attention Aug 24, 2026 Mixture-of-experts LLMs from scratch with OLMoE Aug 06, 2026 Modern LLM architecture and Qwen3 from scratch May 15, 2026 T5 and BART: the encoder–decoder, and a model that writes SQL Apr 28, 2026 GPT from scratch: GPT-2, the KV cache and a model that learns chess Apr 08, 2026 Embeddings and search: Sentence-BERT, cross-encoders and ColBERT Mar 20, 2026 From RoBERTa to ModernBERT: what seven years did to BERT Feb 27, 2026 BERT from scratch: masked language modelling and fine-tuning