AI Coding Agents Are Breaking Big Codebases — Dan Adler, Sourcegraph
Source
youtube.com
Author
AI Engineer
Date
Why it matters
Argues that agents cannot act on code they cannot search, and that AI-generated code accelerates duplication and drift. Sourcegraph's Agentic Batch Changes targets multi-repo changes from a single prompt.
Key takeaways · AI-distilled
Sourcegraph CEO Dan Adler argues the wave of AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition →-written code is making large, decades-old codebases decay through duplicated code, drifting standards, brittle dependencies and new vulnerabilities.
His central claim is that at enterprise scale the bottleneck is context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →, not model quality: agents understand code by searching, so code spread across tens of thousands of repositories stays invisible to them.
Adler frames large-scale code visibility as infrastructure, and presents Agentic Batch Changes as a way to make and audit one change across thousands of repos from a single prompt, with a Mercari case study.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
