Archives

The Case for Calibrated Hybrid Retrieval

PyLate Late Interaction Retrieval

Matryoshka Representation Learning

OpenAI Codex Agentic Coding

Model2Vec Fast Static Text Embeddings

Claude Code CLI Agent

Fine-Tuning Embeddings with Contrastive Learning and MRL

Cosine Similarity Limitations in Vector Search

LLM Inference Became A Systems Problem

Agentic Workflows Need State And Guardrails

Multimodal LLMs Became Product Infrastructure

Small Language Models Found Their Lane

Long Context Is Not A Retrieval Strategy

Structured Outputs Made LLMs Easier To Ship

LLM Evals Became the New Unit Test

QLoRA Made Fine-Tuning Feel Practical

RAG Became the Default LLM Pattern

Instruction Tuning and RLHF Explained

Open-Weight LLMs Changed the Stack