Photo by editor
# Introduction
To create a Document product requirements (PRD) is a common practice. Product management And a common task in fields like software development and the tech industry as a whole. But the story doesn’t end with PRD, and the next big step is turning it into a product, eg a functional software.
This article follows up on that article, in which we turned raw, messy pieces of information into a grounded PRD, and walks you through the same use case (a mobile-friendly app called FloraFriend) to turn that PRD into a working software prototype using Google Antigravity.
Although it is impractical to show the full process of building the software within the scope of the article, we will highlight the key aspects to know about using Antigravity for this purpose, along with some representative excerpts from software developed for illustrative purposes.
# Moving from PRD to software prototype
Get ready, as the process we’re about to describe is where the magic happens. If used correctly Google Anti-Gravity May work partially as Lead Engineer. This is one Downloadable IDE Available for multiple operating systems: You can picture it as a twist on VS Code, such that instead of just typing code, you can handle AI agents that will write it for you.
Needless to say, the first step is to download Antigravity and install it on your machine.
The main element to be aware of in Antigravity is, first of all, its dedication. Agent Manager View: The place where we introduce our software requirements. Then, autonomous agents will plan, implement, and even test the developed solution.
Let’s start by opening it up — I highly recommend picking up “Tokyo Night,” of course, by the way! — theme and, in your local file explorer, create a new, empty project folder, name it flora-friend-app. If you’re familiar with VS Code, you’ll find Antigravity’s UI (user interface) very familiar.
Below you can see the result of opening a newly created folder called “Flora Friend App“, which is currently empty:
First steps with Antigravity
Now comes the interesting part. Open the Agent Manager view by clicking the specified button on the top bar. You’ll see a beginner-friendly explanation of what Agent Manager agents can do:
Welcome screen of Agent Manager
We’ll send this prompt to get started (don’t click the ‘Send’ button yet) that asks for an implementation plan for a mobile-friendly web app:
Work as a Senior Full Stack Engineer. Review the attached PRD for ‘FloraFriend’. Create a comprehensive implementation plan to build it as a mobile-first web app using Next.js, Tailwind CSS, and Shadcn UI. Don’t write the code yet; Rigorously craft the plan artifact first.
This immediate Should be accompanied by a real PRD, for example, as you can find if you follow the previous, related article about NotebookLM for PRD generation. Either way, you have two main options. Attach the PRD: Either paste the code as part of the prompt, right after the request above, or add the PRD file (.docx, .pdf, or similar) to our previously created project folder. Shift + Enter Before sending the full prompt, use the keys on your keyboard to paste the PRD immediately after the request. Big news: Google Gemini LLMs, which fuel Antigravity and its agents, have a larger context window size. In other words, we can add a very long bunch of pasted text to contextualize our prompts in a seamless manner.
After a minute or so, you may get a response that looks like this:
Example output in Antigravity’s agent manager
We can see that a checklist appears on the right side of the Agent Manager window, which lists the software project steps, such as “Start the Next.js app (…)” For example. Here, we have the option to send additional prompts via the “Review” button to review the list and modify items or add new ones, for example, by prompting:
Add “Build Component: Plant Card (with Snooze Logic)” and “Weekender Logic” to backend tasks.
New or modified items will “magically” appear in the checklist.
Let’s go ahead and assume we’re happy with the plan as it is. This is a difficult step, as it is not clear to find the right button to click and move on, especially after several conversations for refinement. The key is to scroll through the artifact boxes generated in the chat stream and find the one labeled “Action Plan”. Open it and, on the right side, click the “Continue” button. This is where the agents start to do their full work, such as installing dependencies, building the database schema through a single file with multiple related tables. schema.sqland so on. Below is an example of the code generated to perform these operations.
Generated schema.sql for our software database.
The entire process is step-by-step, requiring manual review and approval by you before moving on to the next step, but you’ll gradually see how the project planning checklist gets ticked off. Sometimes you’ll need to manually install some dependencies in your PATH or perform similar requirements to let the agents continue their work. But in essence, it just requires careful communication and encouragement!
For a final example, the code snippet for the final app might look like this:
// components/PlantCard.tsx
import { useState } from 'react';
import { Card, Button, Badge } from '@/components/ui';
import { WaterDropIcon, SnoozeIcon } from '@/icons';
interface PlantProps {
name: string;
species: string;
nextWatering: Date;
onSnooze: (id: string) => void;
onWater: (id: string) => void;
}
export default function PlantCard({ name, species, nextWatering, onSnooze, onWater }: PlantProps) {
const isOverdue = new Date() > nextWatering;
return (
{/* Visual Cue for Overdue Plants */}
{isOverdue && (
Thirsty!
)}
{/* The "Matt Feature" - Snooze Button */}
Next scheduled: {nextWatering.toLocaleDateString()}
);
}# wrap up
This article shows how to take advantage of Google’s AntiGravity, one of the most advanced tools available in the Google AI suite, to develop software prototypes. Specifically, we illustrated the logic for converting a PRD into a software prototype.
Iván Palomares Carrascosa He is a leader, author, speaker, and consultant in AI, Machine Learning, Deep Learning and LLMs. He trains and guides others in using AI in the real world.