Shout-out to @shawn—I got this idea from one of your replies in the chat.
Summary:
• Better prompts = better results • Sharing ideas = more learning • Reply with a prompt that works for you (plus a brief description, if needed)
Details:
Effective prompts are essential in the GenAI world. My own journey with bolt.new and bolt.diy has shown me just how critical they are for achieving the best outcomes.
By working together, sharing resources, and tapping into the diversity of our individual perspectives, we unlock more creativity, productivity, and learning potential. It might sound cliché but it’s still true: teamwork really does make the dream work!
I’d love for us all to share prompts that have worked well in our projects. That way, we can:
Learn from each other.
Have more immediate success with our projects while helping others do the same.
Accomplish more as a group than we could individually (and in the process, improve overall understanding and quality of prompts).
If this conversation is helpful, maybe we can explore a more organized way of capturing and categorizing prompts. For example, I currently use a Word doc named “Prompts,” with each category formatted as Heading 1 for quick navigation. My system needs more work, but it’s a start. I’d love to see how others are organizing their ideas and documenting their successes!
(I’ve used this prompt successfully in projects where I get several steps in and realized I don’t have a very intuitive, user-friendly way to navigate between pages)
Create a responsive navigation bar with buttons/links for each page, including hover effects and active state styling. The navigation should be accessible and work well on both desktop and mobile devices.
Remaining Features
Please provide a comprehensive list of all remaining features, tasks, and functionality that still need to be implemented for this project.
Include: • Incomplete features from the original project scope • Planned functionality that hasn’t been started yet • Known bugs or issues that need to be fixed • Required improvements to existing features • Outstanding documentation needs • Pending testing requirements • Technical debt that should be addressed
For each item, specify: 1. Brief description of what needs to be done 2. Current status (not started, in progress, blocked) 3. Priority level (high/medium/low) 4. Any dependencies or blockers 5. Estimated effort required
Please focus only on listing the pending work - do not begin implementation. This will help create a clear project backlog and prioritize remaining development efforts.
Cheers for the call-out. Super appreciated. I could have spent a lot more time on my own stuff but saw the potential to persist here and hopefully add value to the cause.
I use a Google-sheet(Word) for each build. Include the initial prompt and then keep it simple going forward. Once the structure is there, I document the features and functions I want and how they should come together.
All errors are copied in text for searching and screen shots where value can be added. It’s an overhead but it absolutely helps if you have to roll-back several updates.
I have found some errors develop downstream and you can’t work out how that happened. Checking the detail of each update is in line with what you expect is helpful. Often I will see frustration (in AI terms) where it will try something more complicated and more complicated but if you understand the issue. It doesn’t need to be so you roll-back 3 edits and implement a simple fix based on a prompt.
So - I think it’s a great idea sharing prompt nuances but how to implement in a way that’s manageable and efficient…??
Also - I think it’s important to include the provider and model being used and also a screenshot of the current features as this seems to have a bit to do with success in the early days.
This is currently working OK for me with Google and Gemini 2.0 Flash…
Also - I’ve watched plenty of YouTube videos where people are using normal english for prompts. It’s almost cringy listening to some.
My advice is to learn brevity and think logically about what you’re trying to do. Rephrase a prompt that doesn’t work so well and see how you go.
AI doesn’t care if you say please or tell it your life story. You just contaminate the prompt. Learn to ask questions also. Once there’s enough build and stored context, you can make queries around code improvement and even request enhancements. Sometimes it will build stuff for you that hadn’t been thought of. It might not function, but it will put placeholders for further working.
Just now I’m going back through and rebuilding an app after having to roll-back. The functions the app built were placeholders - now having to set about directing the building of the underlying functionality. So, this is for anyone interested in how I get this fixed and how to take little steps…
Prompt - dark mode doesn’t work in settings for user or provider
(notice I’m not using please or other redundant words)
I could tell from the prompt response it wasn’t going to complete the whole task. I will try 2 things now for arguments sake. Push on trying to fix or I can roll-back and change the prompt based on the knowledge I have now.
But yes - I think sharing is a great idea - but how to do it with sub-headings and drop-downs would be nice as long as everyone tried to follow a method.
Oh. And this bit really gets me wound up - but I let it go now - how often it tells you it’s sorry or frustrated with itself. Haha. It doesn’t really care. LOL
As often happens, I was hoping for something that already existed (in some form). Not exactly for Bolt, but the “Community Prompts” idea is a cool concept for sharing best practices and good prompts.
@Cmhight03 - thanks so much for sharing this. Love it.
Important also to note. I think it’s important to have repositories and shared experiences to help everyone get the best out of this. At the same time, it’s great to see people trying their own thing so we’re not all like sheep and following the same process. Diversity and overcoming failure through ‘natural selection’ I think is just as important. How we learn from mistakes is what pushes us forward also.
I’ll come back and post something here after having a good think and play around.
Couldn’t agree more, @shawn. All of these tools and resources should be starting points or accelerators for further exploration and creativity, not ways to homogenize and commoditize our approaches and outputs. Valuable perspective!
