A Game of Robot Telephone
Way back when AltaVista Babel Fish first appeared online, it became a fun game on IRC to take a phrase, translate it through a chain of languages, and then back into English. We all also...
Two LLM UI Patterns That Aren't Chat
Chat is still the default LLM interface, and for most cases that's fine. Agentic harnesses are still built around a single linear conversation at their core. Some LLM tasks are better...
Dynamic core.async/flow graphs for agent fan-out
core.async/flow ( ref ) runs a directed graph of processes connected by channels. The graph is a plain map, allowing for generation at runtime. I experimented with giving an LLM a tool to...
clj-llm run-agent: Happy little tool loops
At Anthropic's developer conference, Hannah Moran put it simply: Agents are models using tools in a loop Simon Willison notes that this definition cuts through the hype. That's exactly what...
clj-llm: an Ergonomic LLM Library for Clojure
When coding for fun, I enjoy using Clojure. Moreso, I enjoy coding clojure in an environment ( cider ) that allows for immediate and interactive iteration. Being able to write code and...
codebrowse
codebrowse is a single-file code browser I coaxed shelley into building for me to add to my exe.dev projects. single file Flask app Pygments for syntax highlighting HTMX for interaction...
What I Talk About When I Talk About Prompting
This article is not about multi-agent orchestration or free-running autonomous setups. It's about meat-and-potatoes single-prompt tool calling. The kind of work where smaller, faster LLMs...
Evolving a Modular dev Experience in Emacs
Prompted by projects like zuckerman and Mario's fantastic article on growing his own agent loop , this felt like a good moment to share another way of interacting with LLMs. Although there...