While the site looks nice and I faancy the idea, I can't help wondering how reliable the information is: Won't these LLMs just create random “facts” if they don't “know” anything about a subject?

I once asked ChatGPT to write a short biography for me and my friend. Little did I know that I was a renowned university professor in history, and ChatGPT even had references for a handful of publications I presumable have written. The biography of my friend was equally absurd.

I believe that was back when GPT-3 hit the streets. Perhaps GPT-4 is able to stick to hard facts?

Another thing, though: The site ought to have a short readme with the relevant context: Obviously, this site is about growing vegetables on the Northern hemisphere. People on the Southern hemisphere should most likely add 6 months to the calendar.

I live in on a smallish island in Denmark at the 55th latitude, and I can definitely not plant out chilli peppers in April — not even in a sunny spot. With some luck I can plant out in mid-May, and I won't be able to harvest anything until at least 100 days later.

I wonder if the data on the site is more suitable for a more southerly continental climate, than the costal climate around here. It would be helpful to know the context.

I built this in a few hours using GPT-4. All the content is AI generated (with one or two minor tweaks by me) and much of the code was written as a collaborative effort between myself and Chat GPT. I basically prompted GPT-4 to dump out vegetable growing information into json files and then built a static site generator that parsed them. This is probably not the most remarkable thing that people are doing with these tools right now, but it blows my mind how much thinking you can offload onto them and still get pretty passable results.