Programming Languages for AI Programing Agents
summary
Argues that languages and tools must expose concrete signals so LLM agents can be grounded: showing that about 5% of get/fetch methods are impure changes generation accuracy, using Codex/ChatGPT plus TypeScript/React demos, deferred blocks and pring-style live examples, unit tests and a run symbolic checker to find failing inputs, and richer data formats (National Weather Service JSON example) and multimodal examples (specs, pre/postconditions, I/O pairs) to iterate and produce reliable, integrable code.
AI-written summary · about
chapters (8)
- 0:00AI programming language research
- 3:15Programming languages in AI world
- 8:34AI agents vs human languages
- 13:12Grounding AI-generated code
- 18:03Data formats for AI collaboration
- 21:13Live programming with AI agents
- 24:02AI expanding programming roles
- 25:27Q&A
related talks
26mLanguage Model Agents Enable Semi-Formal Programming
30mOpen Challenges for Programming Agents
32mProgramming language evolution in the AI era
57mArtificial Languages are Dead. Long Live Artificial Languages!
01h 28mFuture of Software Programming with AI Bots at Hand
46mDemystifying AI Agents
28mAgentic AI
38mCoding Agents & Language Evolution: Navigating Uncharted Waters