What’s new in the last harness — August 2026
A lot has changed in the latest versions of the last harness, my best take on what an AI coding harness should look like.
Automate the boring things by running agent loops at night
As I work on side-projects I tend to accumulate smaller TODOs, like bugs, infra updates, analytics, that I struggle to justify focusing on while I’m directly...
Implementing OpenClaw in Swift and Vapor, part 1: telegram bot & Anthropic APIs
A lot (!) of people are talking about OpenClaw, but not many seem to know how it actually works under the hood.
A simple way to add memory to agents and make them self-improve
LLM agents are great, but them not having memory and not “learning on the job” is less than ideal. Take AGENTS.md, a hack born out this need.
Keeping global commands for CLI agents in sync
Lately, I’ve found myself in this predicament:
Having claude code review your PR
Anthropic recently released a GitHub integration for Claude Code that reviews your PRs. It’s similar to other ones like Copilot, but I’ve found it to be more...