Keynote: Architecting Systems for Agent Experience (AX)
summary
Netlify reports 38% of deploys coming from bots and one internal agent fixes code and merges in 27 seconds, illustrating agent-first flows. The talk shows concrete practices: publish an agents.json manifest with retries, rate limits and least-privilege keys; instrument every agent action with OpenTelemetry traces that trigger contract tests; wrap legacy monoliths with adapter endpoints; and gate self-healing pipelines with policy-as-code and human checkpoints in CI/CD.
AI-written summary · about
chapters (15)
- 0:00Welcome
- 1:53Invisible AI agents rise
- 4:01Zero UI engineering
- 7:14Agent experience defined
- 9:39Human vs agent workflows
- 11:07Expose context with specs
- 13:50Observe in real time with OpenTelemetry
- 15:37Auto-correct in zero UI
- 17:42Testing AI prompts and responses
- 19:29Testers in agentic workflows
- 21:56Pager to autonomous systems
- 23:00Zero UI readiness plan
- 24:48Q&A: AI autonomy vs oversight
- 27:58Q&A: AI exploratory testing
- 30:49Closing remarks
related talks
30mThe Rise of Agent Experience (AX)
19mFrom DX to AX
01mThe Agent Experience (AX): Preparing for the Next 100M Developers
55mKeynote: Exoskeletons, not Autopilots
22mKeynote
19mKeynote: Towards Building Safe and Secure Agentic AI
06mKeynote: Building Agents at Scale
10mKeynote: The Open Source Runway: Building the Foundation for the Agentic Era