Keynote: AI-Powered App Development
summary
Demo shows Copilot CLI adding a new flag and implementation (the --show-models in ~1 minute example) and running parallel reviews across models Opus/Haiku/Gemini; team telemetry reports roughly 100 non-trivial features/week across 10 devs. Workflow uses sub-agents, a skills marketplace (Anthropic/skills), Playwright browser automation for end-to-end validation, and SDKs for node/python/dotnet/go. Raises operational problems: bursty output (six PRs in two hours), review bottlenecks, and drift in developer muscle memory as code generation scales.
AI-written summary · about
chapters (5)
- 0:00Welcome
- 10:07Zero to feature in under a minute
- 22:26Guiding coding agents
- 39:57AI coding challenges
- 52:33Closing thoughts
related talks
32mBeyond the Hype: A Pragmatic Look at How Developers Actually Use AI Tools
55mKeynote: Exoskeletons, not Autopilots
24mThe truth about AI development
47mAI-Powered Software Development From the Trenches
01h 02mPANEL - Unleashing the Future: The Impact of Gen AI on Development Productivity
45mPanel Discussion on AI for Code Generation
48mMastering the AI Shift: Adaptive Strategies for Software Evolution
22mKeynote