You can’t write “AI” without an “I” Control and responsibility in the age of copilots
summary
Built a tiny CLI called lmove that drives vanilla LLMs with structured prompts in YAML/XML and uses prompt-as-spec instead of re-sending prior source. A workflow of small loops plus unit and end‑to‑end tests lets the model generate 20 utility functions, place files by path, and pass automated tests; surprising finding: the prompt and API definition alone often suffice. Rule: never ship code that cannot be debugged locally.
AI-written summary · about
chapters (10)
- 0:00Introduction and purpose
- 1:12My LLM development approach
- 2:39Autopilot vs copilot mindset
- 4:02Human in the loop strategies
- 6:15Aviation checklist analogy
- 8:02Combining human LLM and algorithms
- 9:31Prompting with structured formats
- 11:40Prompt conventions and efficiency
- 13:50Test-driven development revival
- 14:53Debugging and input quality
related talks
45mTest Driven Prompt Programming: How AI is Changing the Way We Code
33mEnsuring quality in an AI-driven world
01h 26mShipping AI That Works: An Evaluation Framework for PMs
37m‘Nurturing’ AI: The Story of How an AI Copilot Was Built
01h 53mAgentic AI: Build a Multi-Agent Application with CrewAI
01h 39mMastering LLMs: From Prompt Engineering to Agentic AI
17mFighting AI with AI
41mAI Driven Mobile Testing