Skill That Survives a Model Swap: Where AI Agent Self-Improvement Should Aim
When an AI agent rewrites its own "working skeleton," how do you tell whether it learned a test-taking trick or real skill? ModularRSI answers by checking whether the gains survive when you swap out the brain model.
Publishing a Subdomain With One Database Row — and the Wildcard DNS Trap
Lay down wildcard DNS and host-based routing once, and spinning up a new subdomain becomes a single database row. But DNS wildcards and TLS wildcards play by different rules, and missing that boundary is how outages happen.
From a Single Sentence to a Shipped Task — How an AI PM Turns Requests into Work
We follow one request — "just build me a company intro page" — from the first conversation through the brief, the deliverable, and the branch point for revisions, to see how an AI PM turns loose words into accountable work.
How Chatbots Fake a Memory
Why chatbots forget your conversations, and how to fix it. We walk through the code for storing conversations safely and keeping them from getting scrambled when you pick up the thread across multiple devices.
Two Weeks Into Meta Muse — Where Is the Personal AI Agent Market Actually Headed?
The next battleground after chatbots is the agent that does the work for you. We dig into Meta Muse's early traction and turbulence, and trace where this market is converging.
The Model Stayed the Same, but Got Better — A Story About Harness Engineering and Loop Design
Squeezing better output quality out of the same model comes down to harness and loop design. Here's the real story of how we rebuilt our own blog pipeline, backed by three recent papers.