AI-First Software Maintenance & AI-First Software Delivery
summary
AI generates code and docs faster than humans can vet, creating an evaluation gap that surfaces as contradictory ADRs, confusing monitors and 45 Jira tickets where 6–10 used to be. The remedy is workflow-level controls: socratic prompting to force human wording, role-based agents (orchestrator, facilitator, specialists), runbooks and living workspace docs, and metrics for harness engineering. A five-hour end‑to‑end mob test runs three times and becomes 13x cheaper by iteration, but change‑failure and time‑to‑restore risk rise without these controls.
AI-written summary · about
chapters (9)
- 0:00AI-first software development
- 3:14Evaluation gap in AI artifacts
- 7:47Socratic prompting to bridge gap
- 10:16AI workflows across teams
- 12:04Human-AI interaction modes
- 15:14AI modes for maintenance
- 17:44Live monitoring with AI
- 22:02Agentic practices & harness engineering
- 24:55Key takeaways & upskilling
related talks
46mRecalibrate - How AI shakes up software team
37mAI-assisted development: Common pitfalls and smarter practices
26mAI-augmented SDLC: Rethinking how we deliver software
31mThe Copilot Wrote It but Who Saves It: Surviving the AI Code Tsunami
18mHumans + AI: The Future of Software Development Teams
28mAI for Everything After Code
27mEngineering in the Age of AI: Making AI Work for Us, Not Replace Us
01mAI Is Changing How We Developer Software!