👋 Welcome! Here are seven simple tips you need to read

1️⃣ Instructions for the First prompt:

• DON’T:
  • Don’t request too many features all at once. Build gradually.
  • Don’t include code
  • Don’t mention programming frameworks: Avoid naming technologies like React, Node.js, Django, Angular, etc. in your product description.
  • Don’t mention specific databases: Avoid naming databases like Supabase, PostgreSQL, MongoDB, Firebase, etc.
• DO:
  • A great prompt includes
    • Description in 1–2 lines (e.g., Barbershop Appointment Scheduling tool).
    • Different user roles (e.g., Two Roles: Barbershop Admin and the Customer).
    • Defines 1 or 2 features per user role, not all at once (e.g., Admin can view their calendar in an admin panel at the page /admin; Client can signup and book appointments). You can introduce additional features gradually later.
    • A paragraph about the design instructions (see examples here)
  • If you have a large document or text with the definition of your project use our helper tool to generate the sequence of prompts: https://pmagent.codapt.app/

2️⃣ Build iteratively: Make sure one feature is working before you jump to the next. If you want to add an integration (to pay, to send emails, etc.) do it an different step.

3️⃣ If you can’t fix an error, click “restore” to break the loop: After a few unsuccessful attempts, go back to the initial prompt that broke the app, click “restore,” and then try different wording or a new approach. The key is to get out of the loop.

Below each message you send to the AI there is a Restore button

4️⃣ Avoid stacking new code on top of layers of “bad code”: If something only works after a few attempts, it’s usually better to restore your project to the state before those failed attempts and then resend the prompt that actually worked. That way, your project stays clean and maintainable. Otherwise all those intermediate steps might have added unnecessary code to your project.

5️⃣ Solid does not have conversation memory: Solid understands your project deeply at every step. But it doesn’t remember past chat messages, by design. So avoid messages like “do it like before” or “it is not working” Just describe what you want, and Solid will handle it!

6️⃣ You can build multiple projects in parallel. You don’t need to keep the window open while the AI is building: just go back to the homepage to start a new project, or even close everything and return later once it’s finished.

7️⃣  If something does not work ask the AI to fix it, be as specific and detailed as possible: Here are a few prompts you can send Solid in case something doesn’t work as expected:

💡 Got an error?

*“when I tried to [describe the action you took] I saw this error: [paste the error message] ”*

💡 See a white screen? (Find logs in “Settings” button)

*“I only see a white screen. Fix this: [paste all logs]”* 

💡 See a “Not found” message? that feature isn’t built yet.

“Build this feature: [describe]”

💡 Something is not working as intented

Be as detailed as possible (which pages, which buttons, what does each button say, what is it doing now, what is it suppose to be doing, etc) 

Examples

Based on the image below, you would send this message:

“When I clicked on “Create Deal” Button at the page /pipeline/deals, i saw this error: { "code": "invalid_string", "validation": "datetime", "message": "Invalid datetime", "path": [ "closeDate"]}]”

⭐️  What’s the tech stack of the web apps generated by Solid?

Solid generates real solid full-stack web apps, built like the real software you actually use every day (Airbnb, Uber, Booking, Dropbox…)—no Supabase shortcuts, no prefab backend tools, no lock-ins. Every part of the stack is written in code, fully customizable, extendable, and production-ready from day one. Build Real Full-Stack Software, the Kind Developers Are Proud to Ship. Stop building toys.

Frontend: