The AI on the other end of the chat box.
A short tour of what's actually happening when you describe what you want. No mystery, no magic. The AI reads your sentence, plans the work, and uses Hatchable to build it.
The mental model
Hatchable runs a small AI agent that knows how to do a fixed set of useful things: make a project, write files, set up a database, deploy what's been written, and check that the result works. Your sentences in the chat box are the only instructions it gets.
Think of it as a competent contractor who has a workshop full of tools they're allowed to use — but won't pick up any tool you didn't ask them to. They write things down, they show their work, and they don't change anything you didn't approve.
What it actually does for you
When you say "build me a chore tracker", here's what's happening behind the scenes — all of it automatic:
Plans
Reads your request, decides what pages, what database tables, and what backend logic are needed. Writes itself a checklist.
Builds
Writes the frontend (the pages people see) and the backend (the parts that talk to the database). Real code, not a wireframe.
Wires it up
Hooks up the database. Adds login if your app needs it. Sets up email sending or scheduling if you mentioned them.
Tests & deploys
Runs its own endpoints with realistic inputs to make sure nothing's broken, then ships the working version to your live URL.
You don't see most of this. You just see "working…" and then a live URL.
How to talk to it well
The single biggest predictor of a good first version is how clearly you describe the use case in the first message. A few habits worth picking up:
Be specific about the people
The AI builds very different apps for different audiences. "For my kids", "for our customers", "just for me" — say it. It changes the login flow, the styling, even the wording on buttons.
Build me a budget app.
A monthly budget tracker just for me. I want to log expenses by category and see what's left in each category. No need for accounts or sharing.
Mention what's important to you
If a feature is the whole reason you're building this, say so. The AI prioritizes accordingly.
A reading log.
A reading log. The thing I care most about is the year-end summary view — I'd like to print it and stick it on the fridge.
Iterate small
"Add a color picker to the new-task form" is way better than "redo the styling and also change the form and add a feature." Small, specific changes give you small, specific updates with predictable results.
Ask, don't assume
If you're not sure whether the AI can do something — just ask. "Can this app send a daily email reminder?" works fine as a question on its own. The AI will say yes/no and tell you what's needed if yes.
When it gets stuck
It will, sometimes. AI is uneven. Here's what to do, in order — most issues clear up by step 2.
- Tell it what's wrong, plainly. "The save button doesn't do anything when I click it." That sentence alone is usually enough. Don't try to diagnose for it; just describe what you see.
- Roll back if a recent change broke things. Every change you ask for is a separate version. If a new one introduced a bug, hit "roll back" to go to the previous version, then try the change again with a clearer description.
- Start the chat fresh on a long-running project. If a conversation gets long enough that the AI is contradicting itself, start a new chat. The project, the database, and the live URL all stay; only the conversation history resets.
- Email support. If you're stuck and the AI can't fix it, support@hatchable.com — include your project name and what's not working. We usually reply same day.
Things to know about working with AI
- It makes mistakes, just like a person. Especially on edge cases or vague descriptions. Always click around the result before sharing it with anyone.
- You're never locked in. Every version is saved. You can roll back any time. You can also export your code and your data if you want to leave Hatchable — see Safety, costs & control.
- It runs inside boundaries. The AI can only do what Hatchable lets it do. It can't email your contacts, it can't read your other projects, it can't spend money you didn't authorize. Each project is its own sandbox.
- It doesn't remember between sessions. Inside a single chat the AI knows what you've said. Start a new chat and you're starting fresh — that's by design, but it means re-stating the project's purpose can help if you've been away.
Most of this is just talking.
You don't need to memorize anything. The AI's good at filling in gaps — your job is mostly to describe what you want clearly and react to what comes back.
Try a build →