Fabian Hildesheim

linkedin · cv

why this site

I recently moved to the US to do AI research at Stanford. I want to build a company. Writing in public creates accountability and forces nuance in articulation — I believe that sharpens thinking. So here we go.

now

Visiting researcher at Stanford HAI, working on AI enterprise adoption with Prof. Brynjolfsson. Technically interested in foundational models, multi-agent systems, and AI adoption. More to come.

general beliefs

I wrote a book on protestant ethics when I was 17. Founded Lupila Learns, an NGO supporting schools in Tanzania with educational infrastructure. Signed the Giving What We Can pledge — 10% of all future earnings. Not because I feel perfectly aligned with the EA community, but because it's the best approximation for moral accountability I see right now.

Also I believe in bias to action and doc-writing.

foundational beliefs

1. Everything is in constant transition, which is a contrast to our state-specific way of perceiving the world.

2. While it is more likely that there is no meaning in anything, we have no other option than acting as if there was one.

3. The best approximation to meaning for me seems to develop a complete and multi-faceted world model, improve my own openness and ability to act upon my beliefs, and with that to pursue happiness for the largest amount of people possible.

thoughts

on AI adoption — march 2026

Since starting my research at Stanford Institute for Human-Centered Artificial Intelligence (HAI), I realized 3 points are not publicly discussed with sufficient clarity:

1. The delta of AI Adoption (whats technically possible vs. whats practically implemented) is more significant than you might think. And this delta is relevant for almost all jobs!

2. There are massive individual differences in productivity uplift. OpenAI calls this "Capability Overhang". They measured across 70+ countries and found that power users leverage 7x more thinking capabilities than average ones. This underscores that using AI is clearly, to some extent, a learnable skill — probably a bit like a mixture of managing a team, formulating a research question and critically reviewing an essay.

3. It's a huge problem that compute, data, AI talent, and funding are already so centralized in the hands of a few people, companies, and countries. Most people on earth have never even used an LLM! Given the 2 points above, it is very important that more people experiment with LLMs, take bolder measures to catch up in adoption, and ensure that productivity gains translate into actual improvements of our society!