from Hacker News

Ask HN: How do you manage your AI prompts?

by siddharthgoel88 on 12/5/24, 6:20 AM with 7 comments

• by runjake on 12/6/24, 10:24 PM

  Depending on the use case and frequency, I either:

  - Save them as a ChatGPT custom GPT or a Claude Project.

  - Create a RayCast AI Command. https://manual.raycast.com/ai

  - Save them as a text snippet in Obsidian notes. https://obsidian.md

• by tobiasnvdw on 12/7/24, 3:27 PM

  Mostly plain text files saved locally for easy copy-pasting.

  I'll occassionally use prompts from the Anthropic library (https://docs.anthropic.com/en/prompt-library/library) and make some minor modifications to them. E.g. I'll modify the "prose polisher" prompt from the prompt library for refining written text in specific ways.

• by cloudking on 12/6/24, 9:30 PM

  For ChatGPT I've found this search extension useful to find previously used prompts: https://chromewebstore.google.com/detail/gpt-search-chat-his...

  Source code: https://github.com/polywock/gpt-search

• by muzani on 12/8/24, 4:34 AM

  I keep it adhoc - models change so frequently that prompts are always broken all the time. Most of the ones I've used last year are no longer relevant.

  "Prompt engineering" may be a thing of the past. These days, you can sketch a vague table on a piece of paper and take a photo of it with a phone, and AI will figure out exactly what you're trying to do.

• by 97-109-107 on 12/6/24, 12:26 PM

  Maybe I'm hijacking, but I see a generalized problem - how do you keep snippets of text that you use in your browser?

  My current kludge is to edit long fields of text in an external editor via a browser addon, and have the editor save all such edits locally.

• by wruza on 12/5/24, 4:30 PM

  I’m thinking of making a simple wrapper around APIs, because web-based AIs tend to dump literal tons of text due to monetary incentives. For now I prepend a standard pseudo-system stub to all my chats, works fine in my case.

• by cyberhunter on 12/5/24, 6:06 PM

  Tired of OpenAI account deletions and Gemini template hiccups? Frustrated with manually typing or copy-pasting prompts every time you switch between LLM clients? If you're like me and want a smoother way to manage your prompts, I built a tool that might be just what you need.

  *The Problem:*

  * OpenAI accounts can be deleted unexpectedly.

  * Gemini templates sometimes fail to work.

  * Re-typing or copy-pasting prompts across multiple clients is tedious.

  *The Solution: DryPrompt*

  DryPrompt lets you create reusable prompt templates with variable fields. You set up the framework once, and then simply fill in the variables to generate the full prompt.

  *How It Works:*

  1. *Go to:* dryprompt.go123.live

  2. *Sign up:* It's free and allows you to sync your prompts across devices.

  3. *Create a template:* Define your prompt structure and mark the parts you want to change with variables.

  4. *Use it:* Copy the template, replace the variables with your specific content, and you've got your ready-to-use prompt!

  *Example:*

  Let's say you need to internationalize multiple code files. With DryPrompt, you can create a template that includes the file code as a variable. Each time, just copy the template, paste in the new file's code, and you'll instantly get the internationalization prompt. No more tedious copying and manual concatenation!

  *Give it a try and make your LLM workflow more efficient:* dryprompt.go123.live