No comments yet.
I'm Tobias, ux software engineer by day and indie hacker by night. The last months I invested a lot of time getting the hang of chatGPT and its predecessor GPT-2/GPT-3.
Based on that knowledge and how to create the best prompts, I created a small copywriting ai tool called schreiberling.app (scribbler in english I guess). The special thing: The user isn't left alone with creating their product copy, they're asked a round of questions to get a better hang of what their product is and who it is for.
This also prompt the user to think more about their product and who their perfect customer actually is. Great for products that haven't launched yet or are not final in their product market fit.
Right now I launched my Beta v2 release, that now uses chatGPT API instead of GPT-3 (davinci), since openAI released the chatGPT api last week.
Some learning going from v1 to v2:
- People didn't know how to provide information without guided questions
- Prompt engineering for multiple languages forced me to create multiple prompts, one for each language
- I created somewhat of a prompt engine to handle things like language, but also things like "tone", to give the user more power to guide the AI to create text that actually fits their needs
- openAIs pricing is amazing. My v1 had GPT-3 calls and 100 people tried it out. I paid $1.69 that month. NOW I have the chatGPT API hooked up which costs 10x LESS than the GPT-3 endpoint. So now I'll pay around $0.17 for 100 user prompts.
- That gives me more freedom to create bigger prompts making them even better without paying myself to bankruptcy
If you have any questions or feedback, I'd love to hear it.
