AI added new ways to search code, but not all of them apply to every problem. Here’s how to choose between Code Search, Deep Search, and MCP.
Check out The Sourcegraph guide to surviving Big Code! In it you'll find 100 use cases for navigating, understanding, and evolving enterprise code at scale.
Sourcegraph and Amp are becoming two separate companies, pushing the frontier of software development in code search and coding agents, respectively. Dan Adler will step into the role of CEO at Sourcegraph. Quinn Slack and Beyang Liu are founding Amp Inc. Our board investors Craft, Redpoint, Sequoia, Goldcrest, and a16z will continue to serve on the boards of both companies.
When you’re fixing a bug, building a feature, or digging into a service you didn’t write, the hardest part is often understanding what the existing code is doing - tracking down definitions, usages, and dependencies across files or repositories. Code Search helps with that every day. But when you don’t know what to search for - or you’re trying to understand how something works, not just where it’s used - Deep Search offers a faster way in. Deep Search lets you ask natural-language questions about your codebase and returns a structured, navigable answer. It uses Sourcegraph’s underlying search and code navigation to explore the code for you, running queries, following references, and summarizing its findings with links, code snippets, and context.
Since the dawn of code, larger dev teams and larger codebases = slower progress. This is why enterprise software development is hard, expensive, and risky. AI can’t automate all dev work, but AI is very good at automating the repetitive parts of coding. This is industrialization: breaking a complex process into 100 smaller tasks, then automating the soul-crushing parts and letting humans focus on the ones they’re good at. By industrializing software development, devs will move faster, not slower, as the codebase and team grow. We believe AI coding agents are best suited to automate the repetitive, mind-numbing parts of enterprise software development, not to try (and fail) to replace humans.
Have you ever wondered how an AI coding assistant works? We’ve been working on our AI coding assistant, Cody, for a while now, and over time we’ve learned that really, context is king.
