Skip to content

Make Proyecta AI More Accurate

Tips to get better results from Proyecta’s AI: start fresh threads, include visual context, and focus on outcomes.

We strongly recommend following these recommendations to get the best results:

  1. Start new threads for each new task: AI performs worse in longer conversations. Use the thread history to create fresh threads for new features or bug fixes.
  2. Include screenshots: Paste or drag screenshots into the chat to show the AI exactly what you’re seeing and what you want changed.
  3. Tell Proyecta the end result you want, not how to get there: Describe your requirements and let the AI work out the implementation.
  4. Use the Visual Editor for small tweaks: For quick text, color, or spacing changes, use the Visual Editor instead of chat.

AI agents perform worse in longer conversations. Old context from earlier requests can confuse the agent and cause mistakes. Start a new thread for unrelated tasks like new features, bug fixes, or changes to different parts of the app.

Click the thread history icon in the chat panel to view previous threads and create new ones.

Proyecta’s AI sees code, not the visual interface you see. Share screenshots to eliminate ambiguity:

  • Paste a screenshot showing the current UI you want to change
  • Annotate in the builder — click the annotate button in the chat to draw arrows, rectangles, circles, and freehand marks directly on a screenshot before sending
  • Describe what’s wrong alongside the image

The AI performs best when you describe the outcome, not the implementation. Don’t over-specify — let it choose the best approach.

We recommend instructions in the format:

users should be able to ...

Examples:

  • “Users should be able to complete checkout in 2 clicks”
  • “Users should see a seamless onboarding experience after login”
  • “Users should be able to easily add or remove contacts”

For quick changes like fixing a typo, adjusting spacing, or changing a color, the Visual Editor is faster and more precise than chat.

  1. Start new threads for each new task
  2. Include screenshots for visual context
  3. Describe the outcome, not the implementation
  4. Use the Visual Editor for small UI tweaks

The AI keeps failing. What should I do?

  1. Start a new thread
  2. Be more specific about the problem
  3. Include a screenshot showing the issue

Can I guarantee the AI’s accuracy?

No. AI is a novel technology and will sometimes make mistakes. The tips above minimize errors but cannot eliminate them.

Any other tricks?

  1. Have Proyecta interview you — ask it to draw out the requirements
  2. Ask Proyecta to stress test your proposal: “What could go wrong with this approach?”
  3. Break large features into smaller, sequential requests