About

I’m Marianne Hovde. I’m a writer, product thinker and compulsive builder who has spent the last few years increasingly interested in one question:

What happens when you stop treating AI as a chatbot and start treating it as part of the system?

I started using large language models for real production work in 2023, including long-form writing, brainstorming and everyday communication. Before long, I was less interested in individual prompts and more interested in making the work repeatable.

That led to structured creative workflows, reusable skills, specialist checking agents, automated writing systems and eventually software built with coding agents such as Codex and Claude Code.

Along the way I’ve experimented with generative video, autonomous creative workflows, multi-model systems, model evaluation, publishing automation and tools built to solve very specific problems — sometimes extremely specific ones.

I’m particularly interested in the messy middle between an impressive AI demo and a system that actually survives contact with real work.

So I test things.

I look at where models succeed, where they fail, what should be automated, where humans still matter, and how changing the surrounding workflow can be more important than changing the prompt.

Some experiments become products or production systems. Some get abandoned. Some produce unexpectedly useful lessons about disappearing horses.

This site is my record of the ones worth talking about.