How to write prompts that get great results

Getting a great app from Baiker is mostly about being specific. Here's what we've learned from thousands of generations.

Lead with the purpose, not the implementation

Bad: "Make a React app with useState and a list component." Good: "A grocery list app where I can add items, check them off, and clear completed items."

Tell Baiker what the app is *for*, not how to build it. The model knows React — your job is to describe the product.

Describe the data

The biggest quality jump comes from describing the data your app works with. Instead of "a music app", try "a music practice tracker where I log sessions by instrument, duration, and notes, and see a weekly summary chart."

Rich data descriptions produce realistic sample data and meaningful interactions.

Name the visual style

Baiker defaults to a clean dark UI. If you want something different, say so: "light mode with pastel colours", "glassmorphism card design", "minimal black and white", "brutalist typography". The model takes style cues seriously.

Iterate with short follow-ups

Once you have a first draft, short follow-ups work better than long reprompts. "Make the sidebar collapsible", "Add an empty state illustration", "Use a table instead of cards for the main list" — each of these is fast and precise.

Use the design panel

For consistent colour palettes and typography, open the design panel (the palette icon in the builder) before generating. Setting a primary colour and font gives the model a constraint to design within, which often produces more cohesive results.