One Feature, Two Timelines: Flying Solo or with an AI Copilot
summary
A case study of a Jetpack Compose "galaxy" UI with custom drawing, many subtle animations and a bespoke collision resolution algorithm that originally costs ~7 months total (≈4 months active work) and ~2,000 LOC for layout. Rebuilding with GitHub Copilot yields the algorithm 90% complete in under five minutes but forces 5–7 days of cleanup for UI recomposition and performance bugs; Copilot helps with docs and simple animations but spits outdated dependency versions and produces fragile code that raises maintenance, context memory, and data‑governance risks.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:45Feature overview
- 8:30Choosing Copilot for testing
- 9:55What Copilot is
- 10:39Rebuilding with Copilot
- 13:27Lessons from Copilot
- 14:30Solo vs Copilot comparison
- 18:47Final impressions
- 19:51When Copilot helps
- 20:34Prompting AI tools
- 21:36AI context concerns
- 22:40AI perception vs reality
- 23:44AI in tech
- 25:12AI governance and sustainability
- 26:54AI limitations
- 27:56Diversity and bias
- 30:06Key takeaways
related talks
34mLessons Learned from Building GitHub Copilot(s)
23mKEYNOTE: From Copilot to Flight Captain
55mCopilot is my Copilot: Pair programming with an AI
05mCopilot, a super quick tour
26mFrom Concept to Prototype
01h 12mJetpack Compose — live coding declarative UI
48mBuild your own copilot with Copilot Studio
09mCopilot App Deep Dive: Intelligence, Context & Agents