So, I was pumped about this AI article thing. I can’t imagine how many clients would find this useful. I even started thinking about the cost I should charge for each article.
I knew I was getting ahead of and in the way of myself.
First, I wanted to have a PRD for the app. So I fired up the ChatGPT o3 mode and started dictating my vision on my trusty voice-to-text app SuperWhisper. I also asked for a step by step prompt that I can feed into Bolt.new.
I immediately thought that I should first create the screen myself which would give me more control over the output. I learnt that from an youtuber. I opened up webcrumbs. It is a prompt to UI design tool. I fired that app and spent nearly 2 hours to perfect the UI.
This is the result.
But I started prompting bolt.new, and I started getting better outputs. Bbetter in the sense that it had its own idea of the app which to some extent was good and modifying them would consume a lot of iterations and credits tokens so I stopped there.
Then I thought creating the article workflow would be better. So I opened up my n8n installation and started stitching together workflow.
I had to spend most of the day to perfect the prompts and the output. I’m about 50% happy with the output.
Next, I’ll have to connect the UI with the n8n workflow. I’ll tackle that tomorrow.
Leave a Reply