Llama.cpp is a powerful and efficient inference framework for running LLaMA models locally on your machine. Unlike other tools such as…

A comprehensive guide covering the local LLM stack from hardware requirements to production deployment. Compare Ollama, LM Studio, llama.cpp and build your first local AI application.

Build with LM Studio's local APIs and SDKs — TypeScript, Python, REST, and OpenAI and Anthropic-compatible endpoints.

Watch now | A new infrastructure stack is forming underneath your AI agents.

Cursor, Claude Code, and OpenAI Codex are forming a composable AI coding stack with orchestration, execution, and review layers instead of consolidating into one tool.

Watch now | The variable that decides who wins the AI infrastructure war isn’t a faster chip or a better model. It’s a compression algorithm.

Why intent and control are becoming the new software architecture

Watch now | Everyone’s talking about the Tamagotchi -- here’s what actually matters.

Credit: Museums Victoria / Unsplash There’s a lot of conversation right now about “context engineering” for dev work; structuring what you feed an LLM so it can do useful things. …

Large Language Models (LLMs) like OpenAI’s GPT-4 have transformed natural language processing, enabling applications ranging from chatbots…

A deep dive into effective caching strategies for building scalable and cost-efficient LLM applications, covering exact key vs. semantic caching, architectural patterns, and practical implementation tips.

In this article, you will learn how to add both exact-match and semantic inference caching to large language model applications to reduce latency and API costs at scale.

A deep dive into exactly how text-only language models are finetuned to *see* images

Persistent AI memory without embeddings, Pinecone, or a PhD in similarity search.

Build custom tokenizers for domain-specific LLMs to improve performance. Learn tokenization techniques, training methods, and implementation steps.

A complete guide to CLAUDE.md, custom commands, skills, agents, and permissions, and how to set them up properly.

We built a 29K-line org AI agent in 4 days with Codex. Here's what broke after launch: credential leaks, silent event-loop deaths, and a teammate who keeps crashing it.

QCon London A member of Anthropic's AI reliability engineering team spoke at QCon London&n ...

Watch now | You’ve seen the videos.

What is this job about really?

The goal of this page is to formalize my answer so that we can judiciously deal with patch requests produced by LLMs.

Implement a coding agent in 131 lines of Python code, and a search agent in 61 lines

AI coding agents burn 20-40% of their context window on codebase exploration. Structured documentation cuts orientation from 20 tool calls to 3.

AI's coding ability is outpacing our ability to wield it effectively. That gap closes in levels — 8 of them. Here's the progression from tab complete to autonomous agent teams.

No longer do we have anyone to question what you’re trying to accomplish.

New research from the University of Waterloo shows that artificial intelligence (AI) still struggles with some basic software development tasks, raising questions about how reliably AI systems can assist ...

The resolution of a human is much higher than we think

When speed matters more than perfection, API documentation can quickly become a bottleneck. In this post, I share how we used thoughtbot’s Claude Skill to generate Postman collections directly from a Rails codebase.

How I used Claude Code to build a GTD-based management system from a borrowed idea, meeting transcripts, and three days of real work.