Google Unveils Code Assist: A New Competitor to GitHub Copilot

At its Cloud Next conference on Tuesday, Google introduced Gemini Code Assist, a sophisticated AI-driven code completion and assistance tool designed specifically for enterprise use. This announcement follows the earlier Duet AI service, which was phased out in late 2023. Google had hinted at a transition from the Codey model to the Gemini framework even before the discontinuation. With Code Assist, Google not only rebrands its previous offering but also delivers significant enhancements.

Demonstrated at the Las Vegas conference, which attracted 30,000 attendees, Code Assist will integrate seamlessly with popular code editors like VS Code and JetBrains. Positioned as a direct competitor to GitHub’s Copilot Enterprise, Code Assist introduces unique features tailored to Google's infrastructure.

One standout feature of Code Assist is its use of Gemini 1.5 Pro, which boasts a one-million-token context window. This capability allows for a deeper understanding of coding context, resulting in more precise code suggestions and allowing for substantial modifications to large code segments. “This upgrade provides the industry’s largest context window at one million tokens, enabling extensive AI-assisted transformations across entire code bases,” explained Brad Calder, Google's VP and GM for its cloud platform and technical infrastructure, during a press briefing prior to the announcement.

Moreover, similar to GitHub Enterprise, Code Assist can be customized to reflect a company’s internal coding practices.

“Customizing code with RAG utilizing Gemini Code Assist has significantly enhanced the quality of coding support for our developers in terms of code completion and generation,” said Kai Du, Director of Engineering at Turing. He anticipates that such customization will significantly boost the overall code acceptance rate. Currently, this functionality is available in preview.

Another feature that sets Code Assist apart is its compatibility with on-premises codebases, as well as those hosted on platforms like GitLab, GitHub, and Atlassian's BitBucket, accommodating hybrid service environments—a distinction not currently offered by Google’s leading competitors in this domain.

In an effort to enrich its capabilities, Google is partnering with several developer-centric organizations, including Stack Overflow, which announced its collaboration with Google Cloud earlier this year. Other notable partners, such as Datadog, Datastax, Elastic, HashiCorp, Neo4j, Pinecone, Redis, Singlestore, and Snyk, will also contribute their knowledge bases to the Gemini platform.

The true measure of success for Code Assist will be developer feedback on its utility and reliability. While Google is positioning itself advantageously with diverse code repository support and an extensive context window, performance issues such as latency or inadequate suggestions could diminish its effectiveness. If Code Assist does not prove to be significantly better than Copilot, a competitor that has already established a foothold, it risks fading into obscurity like AWS’s CodeWhisperer.

In addition to Code Assist, Google announced CodeGemma, a new open model in the Gemma lineup tailored for code generation and assistance, now accessible via Vertex AI.

Google has also unveiled Gemini Cloud Assist, aimed at aiding cloud teams in efficiently managing their application lifecycles. This tool can generate architecture configurations based on user inputs, diagnose issues, identify root causes, and enhance cloud resource management to optimize costs and performance. Cloud Assist will be implemented through a chat interface and integrated into various Google Cloud products.