Comment by emsixteen
3 months ago
I'm on the hunt for ways (system instructions/first message prompts/settings/whatever) to do away with all of the fluffy nonsense in how LLMs 'speak' to you, and instead just make them be concise and matter-of-fact.
fwiw as a regular user I typically interact with LLMs through either:
- aistudio site (adjusting temperature, top-P, system instructions)
- Gemini site/app
- Copilot (workplace)
Any and all advice welcome.
CLI tools are better about this IME. I use one called opencode which is very transparent about their prompts. They vendor the Anthropic prompts from CC; you can just snag them and tweak to your liking.
Unfortunately the “user instructions” a lot of online chat interfaces provide is often deemphasized in the system prompt
ChatGPT nowdays gives the option of choosing your preferred style. I have choosen "robotic" and all the ass kissing instantly stopped. Before that, I always inserted a "be conciseand direct" into the prompt.
i found robotic consistenly underperformed in tasks and it also drastically reduced the temperature, so connecting suggestions and ideas basically disappeared. I just wanted it to not kiss my ass the whole time
Did you made a comparison?
I got did not and also had the impression it performed lower, but it still solved the things I told it to do and I just switched very recently.
If the system prompt is baked in like in Copilot you are just making it more prone to mistakes.