Google.org is providing $1.15 million in funding to App Inventor for a year-long initiative exploring how AI-assisted coding can better support learning.
AI can now write a lot of code for people. That creates new opportunities to build with technology, but it also raises an important question for computing education. When learners use AI to create software, how do we make sure they are still learning how systems work, how to think computationally, and how to solve problems themselves.
With support from Google.org, App Inventor will pursue two related areas of work over the coming year: research into how AI-assisted coding can best support learning and the development of new tools that help people understand and evaluate both existing and newly AI-generated code.
Understanding how AI-assisted coding can support learning
The first project will begin with research involving 1,000 education-sector partners to map how AI-assisted coding is currently being used and perceived. App Inventor will then use workshops with people across computing education and workforce development to build on those findings.
“AI-assisted coding is transforming how people build with technology, but we still have a lot to learn about how to use it effectively when teaching computing,” said Natalie Lao, CEO and Executive Director of App Inventor. “This year, we are launching a research initiative to understand how students, educators, and developers are using and thinking about these tools. Our goal is to co-design an action strategy for AI-assisted pedagogies and scaffolds in computing education that helps us ensure that AI can properly support meaningful learning.”
The project will ultimately produce an action strategy that can help inform the broader computing education community.
Better ways to understand code
The second area of work will focus on the tools people need as AI changes the way software is written.
A major part of that work is Semantic Zoom RS, an AI-assisted approach that can present the same code at different levels of detail depending on what a user is trying to understand. Instead of requiring someone to navigate a single, fixed representation of the program, Semantic Zoom can help surface conceptual, structural, or implementation-level views for tasks such as learning and debugging.
The underlying Semantic Zoom work was presented at the 2026 Blockly Summit, where the MIT App Inventor and App Inventor Foundation teams received the Technical Innovation Award. Over the coming year, App Inventor plans to develop and test the approach, pilot it in real-world settings, and ultimately make the technology available as an open-source Blockly plugin.
“A year ago, most developers would agree with the classic view that ‘code should be written to be read,’ but AI coding assistants can generate enormous amounts of code, making that task close to impossible,” said José Dominguez, Director of Engineering at App Inventor. “Developers need new tools to understand, reason through, and evaluate the code generated, instead of blindly trusting the output from an stochastic system.”
App Inventor will be able to develop and test this work through a global community of more than 30 millions users. The project also aims to extend its impact beyond App Inventor: research findings will be shared with the computing education community, while the open-source Blockly work could make new approaches available to organizations across the broader Blockly ecosystem.
The work began this fall and will continue into summer 2027.