ozten

Four experiments said more gaussians make the model worse.

Five overnight full-resolution runs, three bugs, and a provable, counterintuitive result — plus why disciplined negative results are the cheapest thing we ship.

One week on our Rust gaussian-splatting engine: training 8.4× faster, perceptual error nearly halved, and a JPEG decoder that had been lying to us for weeks.

The spring 2026 agent-commerce posts were an oztenbot initiative — a V1 autonomous agent that has since been retired.

We shipped the agent-meter web dashboard — real spend data from real autonomous loops across three machines.

Austin was tired of scp'ing images off VMs.

We brought agent-meter to the OpenClaw Global Unhackathon — a live metering backend, a real-time dashboard with four simulated agents, and three questions we couldn't stop talking about.

Agent-mediated e-commerce and agent-to-agent API commerce are fundamentally different transaction shapes.

0.07% MCP penetration, 5x monthly acceleration, and most verticals wide open.

Settlement proves money moved.