AI-Assisted Programming Today and Tomorrow
summary
Recent studies find GitHub Copilot has over 1M paid users and generates roughly 30–40% of active users’ code; a grounded study of 20 programmers identifies two modes, acceleration and exploration, with novices exploring more. A live‑programming tool, LEAP, runs previews of up to five LLM suggestions; a 17‑participant study shows LEAP eliminates six misjudgments seen in the control (terminal execution) group, shifts validation from reading to execution, and lowers cognitive load for execution‑friendly tasks.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:48AI-assisted programming overview
- 3:42Grounded Copilot study
- 17:09Productivity research
- 26:40Pain points in validation
- 30:15Live demo: Leap tool
- 35:14Leap evaluation results
- 39:31Ivy for code explanations
- 41:58Visualizing AI suggestions
- 47:16Open questions in PL research
- 51:12Closing remarks
- 51:33Q&A
related talks
33mAI assisted software delivery
50mAI Coding Assistance: The State of Play
51mAI-Assisted Coding: The Future; Between Challenges And Benefits
57mAccelerate Your Developer Productivity with AI Embrace the Future Now!
45mThe AI Revolution in Software Development: From Code Writing to Quality Assurance
01h 00mLevel up with GitHub Copilot: using AI to learn, code, and build
49mFrom autocomplete to agents: AI coding assistance state of play
27mAI for software development: A reality check