Large Language models

OpenAI

GPT-4 is one of the best tools I have used in the last couple of years. It is astonishing technology that makes me more productive.

It is important to have some validation of the output of LLMs, most of what you will get of output will seem reasonable but might be horrendously wrong. This makes it good for programming tasks as it is often easy to get feedback and validate.
The potential for LLMs is quite astonishing. See the [[NLP or LLM]] section for more.

Anthorpic is doing great things.

Standardizing the way you do prompts makes it a lot easier to get the results you want.
Having snippets in Alfred for some standard prompts is something I have experienced a bit with. There are no definite conclusions, but it might be interesting to see the development. Having a defined format for the output would be interesting to see, so you only get the code, formatted data, etc. Alfred also has an excellent workflow for interacting with OpenAI api.
I found recipe searching to be very interesting and fun.
ChatGPT is super handy when it comes to creating shell scripts. super easy to work with.
One of the best innovations to chatGPT was streaming the answer, even though the response comes faster. It gives the user the impression that the AI is thinking and taking time to respond—one of the most fascinating UX choices I have seen.
GPT Monkey: A person who is just pasting results from ChatGPT into the codebase to see if it works. A lovely change from copying results from StackOverflow and seeing if it works. At least know the variables are named within the context of the domain.
LLMs are a commodity.