CI/CD Is Dead, Agents Need Continuous Compute and Computers
summary
Claim: traditional CI/CD breaks at agent scale and must be replaced by a stateful, low-latency agent loop that validates intent continuously. Evidence: thousands of short-lived branches, runner saturation, cache thrash, and a reported fourfold increase in PR volume. Proposed build: Namespace-style hardware+software co-design for a fast cache/orchestration layer that supports no PRs, just intent and plan, routes changes into a premerge layer, lets agents explore parallel commits in a multiverse, and keeps inner-loop builds and validation stateful and sub-minute.
AI-written summary · about
chapters (14)
- 0:00Introduction and speaker bios
- 1:28Why agentic software is breaking traditional CI/CD
- 1:59The fragmented lifecycle of modern software development
- 2:25How traditional CI/CD pipelines work
- 2:55The problems with CI/CD at agent scale
- 4:04Replacing CI/CD with acceleration and orchestration
- 6:12Real-world solutions and the future of agentic loops
- 7:23The role of the human as the agent
- 8:43Why Pull Requests (PRs) are becoming a bottleneck
- 10:00A new architecture: Intent and plan-based development
- 11:58Moving toward fully automated internal/external validation
- 13:46The premerge queue and human-in-the-loop review
- 15:20The future: Parallel development in the multiverse
- 16:51Conclusion: The shifting role of CI and governance
related talks
19mRethinking CI & CD for Agentic Workflows
30mArchitecting Integrity: Responsible Software Evolution in the AI Era
17mCI/CD - Build/Deploy Automation
46mWhen You Run Out of Requirements: What Happens When You Go All-In on Agents
29mOpen-Source Framework for AI-Powered CI/CD Pipelines
33mHumans on the loop, not in it: Taking agentic engineering end to end- 33mHumans on the loop, not in it: Taking agentic engineering end to end
23mCI/CD: Conscious Uncoupling