API Adoption in the Age of AI Code Assistants: Limitations and Opportunities.
summary
Claims that AI assistants hallucinate and produce insecure or incorrect code — a Stanford-style study shows generated code is significantly less secure and another study finds 52% incorrect answers versus Stack Overflow. Demo uses Spotify APIs to auto-generate access and use-case code, actually creating playlists and adding tracks during the demo. Workflow splits into dynamic learning and use-case generation (via an AI co-pilot) and a deterministic code generator for static auth/access code to block fabricated libraries; the hybrid approach reduces hallucinations and works in the presented experiments.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:28Determinism challenges
- 3:13Hallucinations and reliability
- 4:15API adoption importance
- 5:23Key API adoption factors
- 6:27Traditional API adoption steps
- 8:13Demo: guided API walkthrough
- 11:52Limitations of guided walkthroughs
- 12:34AI-enhanced API adoption steps
- 15:27AI vs traditional code generation
- 17:13Fixing hallucinations in AI code
related talks
30mWhat Are the Security Risks Associated with AI-Generated Code? 🤔
39mAI-assisted development: Common pitfalls and smarter practices
41mGoing beyond copilot with AI Agents
33mAI assisted software delivery
46mBreaking AI: Live coding and hacking applicattions with Generative AI
39mBack to Basics: Cutting Through the AI-Generated Code Noise
25mThe role of software design in an AI world
24mAI and Web Development: Hype or Reality