AI inside the IDE
summary
Microsoft uses 3 million JVMs internally and builds a Java Semantic Kernel plus tools for application assessment and diagnostics; code generation shifts review burden because AI writes code that humans must now inspect. Practical solutions include GitHub Copilot CLI with rubber duck multi-model review, automated PR triage via a Tesla plugin that flags human-review-needed sections, and GitHub agentic workflows with JSON “safe outputs” to constrain commits. Copilot alone generates and maintains a Java SDK by weekly PRs that are triaged and merged.
AI-written summary · about
chapters (9)
- 0:00Introduction to AI in IDE
- 1:29Java at Microsoft scale
- 2:33Humans as new bottleneck
- 5:01AI for code review
- 8:51Agentic workflows in GitHub
- 12:45Copilot SDK for Java
- 15:15Why language-specific SDKs matter
- 16:40Reducing human review burden
- 18:05Future of agentic flows
related talks
34mAI and Code Quality: Building a Synergy with Human Intelligence
06mDon’t get one-shotted: Use AI to test, review, merge, and deploy code
58mAI Assistant: Don’t Code Alone
19mWith great power comes great responsibility in code quality
23mAI assisted code reviews Quality and Security at Scale
15mAI's Impact on Development, Post-Production & Support Journeys
41mWill AI Assistant make developers redundant?
44mAI and Code Quality: Building a Synergy with Human Intelligence