Rebuilding CI for AI-Scale Commit Volume — Gautam Korlam, SonarSource

When agents open many PRs, CI and merge queues become the bottleneck. Cache and shard builds, keep failure context across repair loops, and bind approvals to the exact tested commit.

• Korlam's SubmitQueue background: merge queues keep main healthy by testing combinations of changes before they land, but queues sized for human arrival rates stall when agents open many PRs at once.
• He argues agents need durable context from logs and prior failures so a repair loop does not keep rediscovering the same problem, and that overrides and exceptions need reasons and a record that outlives the agent session.
• Approval policy should scale with the change's risk and separate authorship from approval. The talk's volume and latency figures are examples from the speaker's own systems, not general productivity guarantees.

• 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.

• videoPaving the Road to Production: Governed Enterprise AI Agents — Harish Gaggar, Credit Karma
• videoYour AI Isn’t the Problem. Your Infrastructure Is. — Jiachen Jiang, Coder
• videoHow to Actually Tokenmaxx and Save Millions — Alvin Pane, OutRival
• videoI Spent Two Years Building the Median — Hariprasad Sudharshan

• videoAgents, codebases, and teams — Aditya Khandelwal, Amazon AGI LabAI Engineer
• videoAI Writes More PRs. Who Validates Them? — Ali-Reza Adl-Tabatabai, SonarAI Engineer
• videoThe Death of the Code Review: What the Data Actually Says — Laurie Voss, Arize AIAI Engineer

Checking sign-in…

Loading comments…