Founder Notes, 3 November 2026

It is 11:30pm. I did a massive marathon session, and got the bulk of version 1 done. The first agent is in place, and it is working. You can choose from an OpenAI model or an Anthropic model, and within each, the cheapest, middle or most expensive models.

For lead research, the cheap models seem to work fine.

I am fighting the battle of trying to keep everything as simple as possible, but still bringing in the useful but more complex elements of AI, eg memory, context.

I think the fact that the memory and context is AI platform agnostic is good. I also want to eventually allow people to export their memory and context if they wish.

I haven't had a chance to test approvals yet - the times when the model needs a user's input.

Still sometimes in disbelief that with 1 day’s focussed work, I can have a full working product up and running and almost end user ready.

I want to try keep the value offering of this product narrow, and try get a sense early if there is use to others in this product.

Old school style, find the narrowest point of value, and go from there.