How AI is helping me be more optimistic by writing 12,000 pre-vetted questions

Daily Optimist is an app I’ve been working on as a side project on and off for about 6 years.

I was always limited by the amount of questions I could think of though.

When LLMs came out, an obvious choice would be to have an agent generate a question a day.

I think a better choice is to have semi-realtime questions: You invent 50 questions, the AI comes up with 11,950 similar ones (finding similar vector space), and then you pre-vet as many as you can.

If a bad one slips in, you can edit it. You’re not fighting a real-time hydra.

This also works great for code IMO. You need a human to filter the output of the AI.