Cross-platform Strategies for Porting and LLM-Driven Development
summary
Duolingo describes mixing Kotlin Multiplatform (KMP), Rive, Compose Multiplatform and web views: KMP holds stable chess logic (move validation, check detection) and ports a working web board in 1–2 days while iOS build slowdowns are avoided by relinking binaries; Rive powers interactive animations but needs FPS and file-splitting mitigations to avoid ANRs; HTML prototypes and LLMs produce ~90% of a 3D diagram in one shot and LLM-generated grading functions run offline via Jetpack JavaScript engine 1.1.0 for consistent client-side grading.
AI-written summary · about
chapters (12)
- 0:00Cross-platform strategies intro
- 0:44Kotlin Multiplatform overview
- 2:32KMP at Duolingo
- 3:57KMP challenges
- 7:10Rive for cross-platform UIs
- 9:39Rive performance & size
- 12:10Compose Multiplatform UI
- 17:15Web views for content
- 25:57JS without web views
- 31:20AI-assisted porting
- 34:12AI-driven native workflows
- 37:29Summary & future
related talks
38mDesign and Implementation of Next‑Gen App Development Tools Using Compose Multiplatform × AI
41mHow AI Will Change Mobile Development
45mEnhancing LLM Powered Development with Clojure's REPL
34mMulti-coding agent workflows
44mJoint Keynote Address: Accelerating Software Development: The LLM (R)evolution
49mCoding Tomorrow
36mCan LLMs Reproduce Four Years of Compose Migrations? Probing Automation Limits with 279 Real-World XML Projects
52mLivecoding: Creating Beautiful Java Code With LLM and Agents