Ensemble of Machine And Human

Why this exists

In the last week of January 2025, I was burnt out with work, so I did what I usually do to recover: took my workstation to a small cottage in the countryside and built by myself for a week, with no work or family commitments.

Three or four glorious days of working, sleeping, working out, and a wood heated hot tub. Highly recommended for the borderline-introvert's recovery.

I had OpenAI API access, DeepSeek (when it was all the rage), v0, and Windsurf at the ready.

I'm a constant tinkerer - I explore a lot of product ideas, and I kept hitting the same issue using the models: as context grew, focus drifted.

My documentation was patchy. I knew what good looked like, but I did not want to spend all my time documenting everything manually and maintaining an up to date information source.

So I asked:

• Could I build a tool that auto-scaffolds product docs, then keeps them updated as specs evolve using embeddings + RAG to decide which files to change? (No.)
• Could I develop prompting approaches that gave consistent quality across models and model types? (Also no.)

But I did have fun, and I learned a lot.

I got access to the OpenAI beta in May 2021 and, in hindsight, did not do enough with it. I've used Chat since 2023, Codeium/Windsurf and v0 since 2024, and tried multiple agentic approaches through 2025 with mixed results.

I have used AI to design and deliver real tools for internal teams and external clients, and I have presented across the company on AI for product prototyping.

2026 feels different. Agents are more capable than ever, real progress is now possible, and I have too many ideas and not enough execution time.

EMAH is never going to become a product management platform. (Linear can do that.)

It is simply a place to publish projects, experiments, and working notes as I build.