8 entries · newest first
Everything in one stream, newest first: short notes, builder’s logs, artifacts and the things that broke. Bigger projects and research live in things/.
I tested Jev, two Laya checkpoints, Luna at three reasoning efforts, and Sol on the same synthetic roleplay-routing task. High won accuracy. Low missed none of the required recalls. Those are different wins.
Models talk about printable parts fluently and misread their geometry. So my Hermes 3D-print skill measures before it edits, and says what it hasn't checked.
I ran Laya, a small multilingual decision model, on a CPU-only HomeLab box. It was 0.948 sure that a message saying “No refund needed” wanted a refund.
You can't rotate a screenshot of a 3D model. So my agent now sends me a link to the exact revision, and I turn the actual geometry in the browser.
The old page had a video background, analytics and a Twitter widget. Now it is static Hugo with no trackers and no runtime requests.
A self-hosted broker and Hermes plugin for handing an agent a secret without putting it in the chat. MIT, not security-audited.