Marianne Hovde
I build things with AI to find out what actually works.
I experiment with AI-assisted software development, generative media, publishing systems, agentic workflows and whatever else looks interesting enough to pull apart.
Most of my projects start the same way: there’s a problem, a bottleneck, or an idea I want to test. I build something, push it until it breaks, change the approach, and keep going until I either have something useful or know exactly why it didn’t work.
Sometimes that means building software with coding agents. Sometimes it means designing multi-agent workflows, testing models against each other, producing AI video, or discovering that the clever system I built is solving the wrong problem entirely.
This site is where I document the interesting parts.
AI Lab
Experiments, case studies and research notes from working with generative AI in the real world — including the failures, because those are usually where the useful lessons are hiding.
Projects
Tools, systems, products and published work that made it past the experiment stage and became something real.