Google teases Project IDX, an AI-infused code editing thing
Rival to CoPilot and CodeWhisperer sees the Big G join the error-ridden robo-coding market
Google on Tuesday announced Project IDX, an AI-infused cloud-based integrated development environment.
This super IDE "is a browser-based development experience built on Google Cloud and powered by Codey, a foundational AI model trained on code and built on PaLM 2," wrote five Googlers who worked on the project in an announcement. PaLM 2 is one of Google's large language models.
"It’s designed to make it easier to build, manage and deploy full-stack web and multiplatform applications, with popular frameworks and languages," the Googlers – Bre Arder, Kirupa Chinnathambi, Ashwin Raghav Mohan Ganesh, Erin Kidwell, and Roman Nurik – wrote.
IDX is not yet available to the general public – cloud-based here means "vaporware." Google has created a waitlist for interested developers.
Codey was announced back in May at Google I/O, the web titan's annual developer conference. Codey, too, was not available to the public at the time. It subsequently appeared in Google Colab, a hosted Jupyter Notebook service for Python programmers, where it uses AI to suggest appropriate code. Codey also underpins Studio Bot, the AI code helper built into Android Studio, an installable IDE for Android devs.
Google has been here before. In 2010, Mark Miller, an engineer for the corporation, mentioned the Chocolate Factory's internal cloud IDE Brightly in a post to the Dash mailing list. That post eventually spawned the Dart programming language.
But as noted above, Project IDX is different because it has AI, specifically Codey, a text-to-code foundation model that can be embedded in an SDK or application to provide code completion, code generation, and improved code quality – or so it's claimed.
GitHub's rival code assistant Copilot has been found to produce buggy code fairly often (to say nothing about its legal entanglements). Even so, the service is widely used because it can produce helpful suggestions and accelerate projects. OpenAI's ChatGPT also errs a lot when asked to solve programming problems.
Then there's Amazon CodeWhisperer, which produced correct code just 31 percent of the time, according to a study published in April, 2023. ChatGPT and Copilot generated correct code 65 percent and 46 percent of the time, respectively.
Whether Project IDX can live up to that not-very-high standard won't be clear until public availability.
The Googlers involved in the project sound optimistic that the company's AI can generate worthy code.
"With Project IDX, we’re exploring how Google’s innovations in AI — including the Codey and PaLM 2 models powering Studio Bot in Android Studio, Duet in Google Cloud and more – can help you not only write code faster, but also write higher-quality code," they said, even as they acknowledged there's room for further improvement.
Perhaps the project's most compelling selling point is its promise to generate preview builds for multiple platforms – initially for the web, and eventually Android, and iOS. And it may win points for familiarity – Project IDX is based on Code OSS, the foundation for Microsoft's free and widely used Visual Studio Code.
There's also something to be said for being able to drop into a development environment with a browser, from any device. Of course, there are already many such services, though perhaps without as much AI special sauce.
One potential downside is that Project IDX makes Google Cloud a dependency, and it would not be surprising were Google Cloud to look for opportunities to upsell Project IDX users on adjacent services.
Pricing for Project IDX is yet to be disclosed. ®