Figma Make-a-thon winners share the prompting strategies that worked
Figma's first global Make-a-thon challenged the community to see what could be built with Figma Make, the company's AI-powered tool for creating interactive prototypes. More than 10,000 creators participated, competing for $100,000 in prizes. Entries ranged from a Severance-themed virtual escape room to a racing game, but the winning projects share a common thread: they all demonstrate effective techniques for turning rough ideas into polished, working prototypes through smart prompting and iterative workflows.
We looked at the top projects to extract the practical lessons on how to approach a Figma Make build from scratch.
Map out the app's architecture before writing a single prompt
Sebastian Lumbi, who took second place with his "Frame by frame SVG animator," shows that understanding the underlying structure of an app pays off even when the AI writes the code. The tool lets users paste SVGs from Figma to create handcrafted animations and export them as GIFs.
Lumbi began by researching key technical decisions—specifically state management and component interaction within Figma Make—before starting an iterative conversation with the AI. He admits he didn't have a detailed prompting strategy at first, but found that clearly communicating how he wanted the app's architecture to be laid out made debugging far easier later on.
Takeaway: You don't need to be a developer to benefit from basic technical knowledge. Telling the AI how you want the app structured, or asking another AI for recommendations on architecture, makes the output easier to maintain and fix.
"If you're struggling to fix a bug or your app has crashed, it's often more effective to go back to a previous point and try a different approach from there."
Lumbi also recommends leaning on the revert option as a safety net, claiming it removes the risk of experimenting with a new feature. His starting prompt was straightforward:
Use the design to create a frame-by-frame animation creator/player. The top part is the canvas where I can draw the frame, and the bottom part is where I can see the frames and create new ones.
Polish your design in Figma before entering Make
Grand prize winner Cara Ellis took the opposite approach to Lumbi, spending significant time on the design phase before ever opening Figma Make. Her "Web poetry" project lets users draw a bespoke alphabet, then type a poem using their custom font.
Ellis sketched her vision on paper, recreated it in Figma Design, and only then moved into Figma Make. That up-front work meant she needed fewer prompts to refine the final product—just 90 in total—and the AI reproduced her detailed vision on the first pass.
Takeaway: Design first, build second. Figma Make responds best when it has a clear visual target, which saves prompt iterations down the line.
Top tip: Ellis encourages relentless testing: "Test your app as many times as you possibly can, and then get your friends to test it, too. Every time something is even slightly off, tweak it."
Prompt for modular code to simplify editing
Susanna Nevalainen's "Moxy" 3D configurator won best prompt. The tool lets users upload custom branding into mockups, a solution aimed at helping freelance designers showcase their work.
Nevalainen worked from a layout in Figma, then used sketches to communicate visual details that were hard to describe in words—such as resize and move handles. Her key move, though, was prompting the AI to organize the generated code into smaller components, separate files, and logical folders. This modular structure enabled far easier ongoing refinements. She also instructed the AI to ask clarifying questions when it was uncertain about her intent, a practice that saved her from unnecessary rollbacks.
Takeaway: Figma Make generates code that can be downloaded and edited in any external editor. Prompting for a sensible code structure from the start makes that code useful beyond the initial prototype.
Top tip: Nevalainen sees Figma Make as a tool that bridges design, development, and AI—useful to professional developers who might otherwise view it as a no-code-only feature.
Start in the Figma Make chat loop to unlock API integrations
Matt Schroeter's "Weather Synthesizer"—named most creative—turns live weather forecasts into musical compositions. He didn't start from a design file at all, choosing instead to co-create entirely through the Make interface.
The AI suggested improvements that Schroeter combined with his own ideas through roughly 150 prompts. The experiment proved how far the tool can go: by using API integrations for the audio engine and weather data, he built something functionality rich for native browser interaction.
Takeaway: Figma Make isn't limited to static UI design. It can work with real data sources and external services to produce fully functioning interactive applications.
His starting prompt was creative in its constraints and specificity:
A synthesizer in the shape of a plant that reacts to changing environmental/weather conditions: the current weather, wind strength, temperature, and air quality in Vancouver, Canada.
Use Make to test assumptions during the design phase
Johannes Specht's third-place "Plan That Trip. Now" helps groups settle on destination options. Rather than alternating between Figma and Make, Specht built high-fidelity prototypes directly, then handed those files to the AI and added base logic through prompts.
He used live browser testing through the tool to refine layouts in real time, essentially treating Figma Make as a dynamic validation environment during the design process. With roughly 250 prompts, he iterated on responsive behavior across screen sizes without leaving the Make environment.
Takeaway: Figma Make lowers the cost of prototyping so you can uncover pain points and test flows in a realistic environment—before diving into a full design effort.
Short instructions are a testable starting point
Daniella Marynova and Max Pradella's "Package Customizer," the winner for most innovative use, began with a well-organized Figma file complete with tokens, components, and clear frame names. The duo, inspired by Pradella's own experience at a delivery company, set out to build a full parcel configurator with an augmented reality view to preview packages on a desk.
For a project of that complexity, they started with a single, minimal prompt.
Create a parcel configurator.
They kept prompt iterations around 450 total, but followed a strict methodology of short, focused instructions. Marynova praised this approach as less likely to confuse the model and claimed it contributes to easier testing and faster identification of issues.
Takeaway: Short prompts are reliable. The team also maintained a log of prompt versions—a simple practice that saved time whenever they needed to roll back to a working state.
"Short prompts work better, are easier to test, and less likely to confuse the model. We also kept a list of prompt versions—it saves time if you need to go back."
The unexpected themes across the winning projects: Figma Make rewards detailed up-front thinking, rewards users who understand basic app architecture, and performs surprisingly well in complex, data-rich domains as long as each user treats it as a conversation partner.
