LLMs

I like cooking recipes. They’re clean, serene documents where, at some point, one or more utensils enter the stage to perform a task. If you were to write one on how to use AI to augment your work as a technical writer, though, you would have a hard time deciding when and where LLMs come out of the toolbox to aid you. The variety of flavors in which AI presents itself doesn’t help. This is why I’ve come up with a framework that describes applications of the different AI tools at our disposal.

A universal solution to better chatbot responses

How to ensure your coding agent has the same context as you

Learn how to steer an LLM step by step instead of hoping a single prompt gets it right.

Retrieval-augmented generation makes language models far more useful by grounding them in real data, But it also raises a hard question: who is allowed to see what? This post shows how authorization can be enforced before retrieval, ensuring that RAG systems remain powerful without becoming dangerous.

It's a seductive distraction from the advances in AI that are most likely to improve or even save your life

Learn about AI agents, automation, productivity, and how to get more done.

Meet DoAnything, the AI agent that handles your most stressful tasks. Start a business, plan events, apply to jobs, and more.

Weeknotes 371 - The real shift is not from human coders to AI agents—it's from coding to engineering the environment where agents are co-designers. And other news on AI companion devices and robots as CES.

Happy New Year, and Welcome to Gas Town!

It has been three days since I launched Gas Town! 🔥⛽💥🛢️🔥 Woohoo!

Manus is the action engine that goes beyond answers to execute tasks, automate workflows, and extend your human reach.

Lovable is an AI app and website builder that turns ideas into production-ready software without coding. Launch faster with AI.

Deploy fleets of AI coding agents with the world's best UX for managing and collaborating on agent workloads. Fully open source CLI and local desktop orchestration, or use cloud sync and remote agents to scale to your whole team and all your devices.

A hands-on exploration of OpenCode, an open-source coding agent built with a client/server architecture. Learn how AI tools, LLMs, and real-world constraints come together to create a powerful developer experience.

If you're building AI agents, you're probably sitting on a lot of untapped data. Experts share how to put it to use.

Agent OS is a popular system spec-driven development with AI coding agents. It's a free open-source framework from Builder Methods, created by Brian Casel.

Design OS is a product planning and design tool that guides you from product vision to production-ready components. Free and open source.

Kagi Search Help

Whenever a new pull request is submitted to our agent’s GitHub repository, we run a bunch of CI/CD operations on it. We run an opinionated linter, we run typechecking, and we run a bunch of unittests. All of these work well, but none of them test entire workflows end-to-end. For that end-to-end testing, we introduced an eval pipeline. This is part of the Building an internal agent series. Why evals matter The harnesses that run agents have a lot of interesting nuance, but they’re generally pretty simple: some virtual file management, some tool invocation, and some context window management. However, it’s very easy to create prompts that don’t work well, despite the correctness of all the underlying pieces. Evals are one tool to solve that, exercising your prompts and tools together and grading the results.

Agents are extremely impressive, but they also introduce a lot of non-determinism, and non-determinism means sometimes weird things happen. To combat that, we’ve needed to instrument our workflows to make it possible to debug why things are going wrong. This is part of the Building an internal agent series. Why logging matters Whenever an agent does something sub-optimal, folks flag it as a bug. Often, the “bug” is ambiguity in the prompt that led to sub-optimal tool usage. That makes me feel better, but it doesn’t make the folks relying on these tools feel any better: they just expect the tools to work.

Retrieval-augmented generation breaks at scale because organizations treat it like an LLM feature rather than a platform discipline. Enterprises that succeed with RAG rely on a layered architecture.

Yes, this is the umpteenth article about AI and coding that you’ve seen this year. Welcome to 2025. Some people really find LLMs distasteful, and if that’s you, then I would recommend t…

For IT and HR teams, SLMs can reduce the burden of repetitive tasks by automating ticket handling, routing, and approvals, while providing substantial cost savings versus LLMs.

If your access control lives in a prompt, it isn’t access control. Authorization decisions must be deterministic and enforced before an LLM ever sees data. Treating AI as a policy engine is a category error with real consequences.

Anthropic installed an AI-powered vending machine in the WSJ office. The LLM, named Claudius, was responsible for autonomously purchasing inventory from wholesalers, setting prices, tracking inventory, and generating a profit. The newsroom’s journ

GitHub release .NET Native AOT A local-first document summarization tool that uses LLMs (via Ollama), vector search (Qdrant), and document conversion...

The basic concept of human intelligence entails self-awareness alongside the ability to reason and apply logic to one’s actions and daily life. Despite the very fuzzy definition of ‘hum…