

Building CI for the age of AI Agents with Aayush Shah
Jul 22, 2025
Aayush Shah, co-founder of Blacksmith and a systems expert from Cockroach Labs and Faire, dives into the evolving world of CI in the age of AI. He discusses how the surge of AI agents is reshaping CI workloads and the need for improved visibility in testing. Aayush shares insights on their innovative approach to cloud economics, revealing how traditional storage isn't cutting it for their unique demands. He also highlights Blacksmith's capabilities in handling multi-tenancy and resource allocation, all while navigating the challenges of continuous integration.
AI Snips
Chapters
Transcript
Episode notes
GitHub Actions' Incentive Misalignment
- GitHub Actions pricing relies on per-minute charges, discouraging speed improvements that reduce usage.
- GitHub treats Actions more as workflow orchestration lacking CI observability features, leaving gaps others can fill.
Manage Flaky Tests Proactively
- Identify flaky tests by tracking failure patterns and stress new tests before merging.
- Prioritize investigating failures relevant to current changes to avoid wasting developer time on unrelated flaky tests.
CI Workloads Suit Dedicated Hardware
- CI workloads do not require running inside customers' cloud accounts and are location-agnostic.
- This enables Blacksmith to run on cheaper, dedicated hardware outside hyperscalers, optimizing costs and performance.