A review of dynamic documents and AI - specifically LLMs - and their role in end-user programming, including the importance of differentiating between a tool and a machine.
Tools should offer human-in-the-loop AI amplification, whereas machines automate human tasks.
There are two main ways to use LLM: giving instructions to an AI and letting it directly do stuff; or having it spit out code that can be run by the user.
It is important to differentiate the two, as the former enables soft reasoning that is basically impossible to do in traditional code, while the latter is slower and more expensive.
Reliability will improve in the near future, and it will be interesting to see how programming languages evolve to make this process easier.
Overall, the author is interested in AI-enriched tools that allow humans to think and work in their domain. Fuzzy reasoning and direct human involvement are essential.
As the author is interested in this topic, they will share more soon and hopes to hear from readers on the topic.