In our latest live Q&A session, Sourcegraph co-founders Quinn Slack and Beyang Liu, along with the DevRel team, Ado Kukic, and YK Sugi, shared exciting updates and engaged with our amazing community. The session highlighted Cody's evolution and our commitment to building the most informed AI coding assistant for real-world development scenarios.

Introducing Deep Cody: Enhanced Context Understanding

The highlight of the session was YK's demonstration of Deep Cody, our experimental feature that represents a significant leap forward in AI-assisted coding. Deep Cody builds upon Cody's core strength – context understanding – by introducing advanced capabilities that allow it to:

• Access additional context beyond the immediate codebase
• Make informed decisions about what context it needs
• Perform multiple context-gathering iterations

What sets Deep Cody apart is its ability to solve entire classes of previously challenging problems. From analyzing repository statistics to accessing GitHub issues, Deep Cody demonstrates a more comprehensive understanding of your development environment.

New Features and Improvements

Visual Studio Support

We're excited to announce Cody support for Visual Studio is now available in preview for Windows users. This release continues our commitment to meeting developers where they work, regardless of their preferred IDE.

Prompt Library

The new prompts interface introduces shareable prompts, allowing teams to:

• Create and share standardized approaches to common tasks
• Maintain consistency in code generation across teams
• Build upon successful prompting strategies
• Include context and command execution capabilities

Context Improvements

We've enhanced how Cody handles context with:

• Improved embeddings-based retrieval
• Manual context editing capabilities
• OpenCTX for integrating various data sources

Focus on Real-World Development

A key theme throughout the session was our dedication to solving real-world development challenges. As Quinn emphasized, "99% of software development happens in existing complex codebases." While others focus on creating new applications or flashy demos, we're committed to helping developers work effectively with legacy code and complex systems.

This focus resonates with our community, as evidenced by user Michael Henderson's comment:

Cody fills my need to understand legacy code. Copilot doesn't seem to consider that to be a use-case.

Looking Ahead

Our roadmap includes several exciting developments:

1. Code Review Integration: Expanding Cody's capabilities through the Cody API
2. Enhanced Context Architecture: Working on improved caching and context handling
3. IDE Support: Continued development of plugins for various editors
4. Performance Improvements: Focus on speed and reliability of features like Smart Apply

Get Involved

We're actively seeking community feedback and contributions. You can:

• Join our Discord community
• Help others in our Community Forum
• Follow @Sourcegraph and @SourcegraphCody on X
• Provide a review for VS Code or JetBrains in their Marketplaces

Conclusion

This Q&A session showcased not just our technical progress but also the strength of our community. As we continue to evolve Cody, our focus remains on building tools that solve real problems for developers working in complex, existing codebases.

Whether you're maintaining legacy systems, working on large-scale applications, or simply looking to improve your development workflow, Cody is being built with your real-world needs in mind. We're excited to continue this journey with our community and look forward to sharing more updates in our next session.

---

Want to try Cody? Visit sourcegraph.com/cody to get started. For the latest updates, follow us on X/Twitter.