Sourcegraph CEO: AI-generated code is decaying large codebases at scale
AI Engineer · youtube · 2026-10-05
At AI Engineer World's Fair 2026, Sourcegraph CEO Dan Adler argued that owners of huge, decades-old codebases (banks, cars, airlines) face accelerating decay — duplicated code, drifting standards, brittle dependencies, and new vulnerabilities — as coding agents produce a tidal wave of AI-generated code.
Key points:
- Agents understand code by searching, but "you can't grep what you can't see": at enterprise scale, context and code visibility — not model quality — are the bottleneck.
- Adler argues large-scale code visibility is infrastructure, and introduced Sourcegraph's Agentic Batch Changes, which makes and audits changes across thousands of repos from a single prompt.
- Includes a case study with Mercari and a call to dev tools builders.
More from coding & agent
- Building a personal AI pipeline to summarize and query expert content: yt-dlp to SQLite — QuestionAsker2030 · 2026-10-05
- Redditor drafts yt-dlp + Whisper + SQLite pipeline to auto-summarize expert content — QuestionAsker2030 · 2026-10-05
- SpaceX engineer: AI agents now merge PRs on their own — over 1,000 a month — itsOmSarraf_ · 2026-10-05
- Opus 5.5 as daily driver, GLM for cheap prod: a practical model split — ivan_bezdomny · 2026-10-05
- Astra agent reverse-engineers macOS binaries to clone all 27 Liquid Glass effects in GPUI — steipete · 2026-10-05
- Local MCP server gives coding agents a project map: 9.3x fewer tokens on a real 12-step task — Honest_Traffic_8613 · 2026-10-05