Google upgrades its programming agent Code Assist with Gemini 2.0, adds source integrations

After releasing new generative artificial intelligence model, Google Updated its coding assistance tools for use Gemini 2.0 and expanded its connected external sources.

Code Assist will now run on the recently released Gemini 2.0, providing a larger contextual window into an enterprise’s larger code base.

Google will also launch the Gemini Code Assist tool in private preview. The platform will connect to sources such as GitLab, GitHub, Google Docs, Sentry.io, Atlassian and Snyk. This will allow developers and other coders to seek assistance from Code Assist directly within their IDE. Previously, Code Assist connected to VS Code and JetBrains.

Ryan J. Salva, senior director of product management at Google Cloud, said in an interview with VentureBeat that the idea is to allow programmers to add more context to their work without interrupting the workflow. Salva said that Google will add more partners in the future.

Code Assist was formerly known as Duet AI Launched for enterprises in October. As organizations look for ways to streamline coding projects, demand for AI coding platforms like GitHub Copilot continues to grow. Code Assist adds enterprise-grade security and legal compensation when it launches as an enterprise option.

Artificial intelligence at work for developers

Salva said connecting Code Assist to other tools used by developers can provide more context for their work without having to open multiple windows at once.

“Developers use a lot of other tools throughout their day,” Salva said. “They might be using GitHub, Atlassian Jira, DataDog, Snyk or all these other tools. What we want to do is give developers the ability to bring additional context into their IDE.

Salva said developers only need to open a Code Assist chat window and ask it to summarize the latest comments on a specific issue or the latest pull request on the repository, “so that it can query the source and bring the context back to the IDE, [the] Large language models can synthesize it.

AI coding assistants were one of the first important use cases for generative AI, especially after software developers started using ChatGPT to help with coding. Since then, a series of enterprise-focused coding assistants have been released. GitHub releases Copilot Enterprise February, and Oracle launched its Java and SQL Coding Assistant. harness Launched a coding assistant built with Gemini Provide instant advice.

at the same time, Open artificial intelligence Anthropic began to provide interface features that allow programmers to work directly on their chat platform. ChatGPT canvas Allows users to generate and edit code without copying and pasting it elsewhere. Open artificial intelligence also Added tool integration For example, VS Code, XCode, Terminal and iTerm 2 in the ChatGPT MacOS desktop application. at the same time, Anthropic selection roll out Claude’s artifacts So Claude users can generate, edit and run code.

Not Jules

Salva pointed out that although Code Assist now supports Gemini 2.0, it is still fully separated from julesthe coding tool Google announced during the launch of new Gemini models.

“Jules is actually one of many experiments launched by the Google Labs team to show how we can use autonomous or semi-autonomous agents to automate the coding process,” Salva said. “You can expect that over time , experiments graduated from Google Labs, these same capabilities may become part of products such as Gemini Code Assist.”

He added that his team works closely with the Jules team and is excited to see Jules’ progress, but Code Assist remains the only generally available enterprise-level coding tool powered by Gemini.

Salva said early feedback from Code Assist and Jules users has shown significant interest in Gemini 2.0’s latency improvements.

“When you’re sitting there trying to code and trying to stay in a flow state, you want these types of responses to come within milliseconds. Any time developers feel like they’re waiting on the tool is a bad thing, so what we get from that is The responses are getting faster and faster,” he said.

Coding assistants remain critical to the field of generative artificial intelligence, but Salva said the way companies develop code-generating models and applications may change in the coming years.

Salva pointed out Accelerating DevOps State of 2024 Report A survey from Google’s DevOps research and assessment team showed that 39% of respondents do not trust code generated by artificial intelligence, and the quality of documents and delivery has declined.

“Over the past four years, as an industry with AI-assisted tools, we have focused primarily on throughput productivity improvements and speed improvements,” Salva said. “When we started to see that this was associated with an overall decrease in stability, I was skeptical. The discussion next year will really turn to how we use AI to improve quality across multiple dimensions.”