About DomGPT

Who am I?

I live in the UK, and have worked in finance, as an entrepreneur and in artificial intelligence. I'm not keen on sharing more info - for the reasons best given by Gwern here.

I am not affiliated with Dominic Cummings and have never met him. If you are him, please email (domgptlarp at Google's email service)- I'm keen to clear up any IP issues / donate to Maths Circles / whatever.

What is next?

The site is based on Cummings' blog posts. I'm keen to incorporate his tweets and podcast / youtube interviews. These cost money directly and time through opportunity costs. If I get a reasonable number of coffees then I'll implement them in the next few weeks. Including these materials would improve the quality of the output.

Longer term I'd also be interested in making similar sites for other figures. Recently I have been thinking about Michael Nielsen and Stafford Beer.

How is the site made?

The idea for this site is mine, but the tools recommended below (as well as all the code) came from Claude. I have no technical ability. It was partly inspired by a comment from Tyler Cowen: we should be shifting to a more question-lead paradigm of learning.

The full stack is:

1. Pinecone and Cohere worked together to encapsulate and represent Dom's thoughts in a way which can be read by computers.
2. GitHub is a kind of exercise book used by me and Claude to communicate with Railway and Vercel below.
3. Railway does the heavy lifting - 'backend', while Vercel does the presentation - 'frontend'.
4. Claude takes the references generated by Pinecone and Cohere and presents it back to you the user.
5. Grok / Aurora generated the background images.

I'd be open to making other pages if you want a similar site made about you or someone you think highly of. An ethical screen exists ('No' to Greta or Andrew Tate).