Greptile Enhances Bug Detection with GPT-6 Astra and Other AI Tools

Greptile builds AI agents that review and test code changes with a comprehensive understanding of the codebase. “We are in the business of reducing bugs and making the world’s software more reliable,” says Co-founder and CEO Daksh Gupta.

To achieve this mission, testing an application’s behavior is as crucial as analyzing its code. GPT-6 Astra provides Greptile with a means to automate testing at scale.

Testing Applications at Scale with GPT-6 Astra

Validating a pull request requires checking the logical correctness of the code, ensuring it meets the developer’s intentions, confirming security, and preventing disruptions to existing functionality. Some of this testing necessitates running the application, as code review alone may not uncover issues.

For instance, Daksh mentions a scenario where a button disappears when a website is accessed on an iPhone. With GPT-6 Astra, Greptile can initiate a development server for a pull request, adjust the browser to an iPhone-sized viewport, and simulate user navigation through the application. The model’s visual capabilities allow it to inspect the screen and identify the hidden button.

“A QA firm testing all of your changes is just impractically slow considering how fast we’re able to ship code with these models,” he states. GPT-6 Astra enables Greptile to automate application testing effectively.

Strengthening Customers’ Defenses with Daybreak

According to Daksh, more advanced models also facilitate exploitation by attackers, increasing risks for companies, including those that were not typical targets before.

“We want to make it very easy for even a startup to secure their systems,” Daksh explains. “With Daybreak, we’re able to do this in a way that is very safe, ensuring our customers have more sophisticated defense mechanisms than the adversaries have offense mechanisms.”

Reducing Bugs and Rework with Codex

Greptile’s Knowledge Base catalogs a company’s codebase and documents its functionality in a format accessible to both people and agents. This documentation is made available to Codex through a Model Context Protocol (MCP) server.

This connection allows Codex to generate code that aligns with the customer’s existing patterns. Daksh notes that thousands of customers have linked Codex to Greptile’s Knowledge Base.

To assess the impact, the team monitors how often AI-generated code is reverted. “When we provide agents with context, fewer reverts happen, and customers save time and inference,” Daksh says. These savings can be redirected towards new projects instead of reworking existing changes.

Greptile also utilizes Codex to develop its own software and agents. “We are massive users of Codex,” Daksh adds. “We have enormous amounts of usage across every single function, engineering in particular.”

Looking Ahead to Faster Development with Ultrafast

For Greptile, Daksh emphasizes that the primary bottleneck is the speed at which the team can transform ideas into functional software. He believes model intelligence is already adequate for many tasks, viewing Ultrafast as a means to accelerate development.

His long-term objective is to simplify the process of modifying and deploying critical enterprise software to the ease of shipping a side project, with OpenAI assisting Greptile in building the necessary validation.

Source: OpenAI News