DevRelling in the age of GenAI
summary
Live demos use Anti-gravity LLM plus reusable skills to generate and iterate Python extractors for large files (a VCV Rack patch and a 15k‑line schematic), exposing two concrete failures in LLM output (an inverted diode and incorrect BAT54S representation) and reducing analysis time from 25s→11s. The workflow requires 4–5 refinement loops, testing generated code, and a shift to test-and-verify rather than blind trust.
AI-written summary · about
chapters (11)
- 0:00Welcome
- 1:28Coding to learning with AI
- 2:11Locus of control in AI
- 3:59AI content limitations demo
- 8:18Verify AI results
- 10:03Skills for better LLM performance
- 16:24Why skills improve LLM
- 17:27AI analyzing modular patch demo
- 23:22Managing AI context length
- 27:38Improving AI-generated code
- 28:40Ownership of AI content
related talks
43mAI Development Lifecycle: Learnings of What Changed with LLMs
38mNobody Pages the LLM: Engineering Rigour for Vibe Coding
01h 00mMalignant Intelligence: Prompt engineering and software archeology
24mNatural Language Coding Revolutionising Programming with GenAI
31mRethinking DevOps and DevSecOps with Generative AI: Insights
15mChallenges in moving Gen AI apps to production
48mGoing Beyond Prompt Engineering
12mAI-generated recap of Devoxx Belgium 2024 opening keynote : Genesis