O11ycast Ep. #87, Augmented Coding Patterns with Lada Kesseler
13 snips
Dec 23, 2025 Lada Kesseler, a seasoned software developer known for her expertise in extreme programming and AI-assisted coding, joins the conversation to explore navigating the complexities of AI coding assistants. They delve into the AI black box phenomenon, discuss semantic zooming for clarity, and emphasize the importance of high-level tests and feedback loops in managing agent behavior. Lada shares her successful experiences with AI in frontend development and offers practical tips for maintaining relevance in AI-generated documentation while ensuring effective collaboration.
AI Snips
Chapters
Books
Transcript
Episode notes
Semantic Zoom Controls The Noise
- AI coding assistants are a silent black box with invisible context that makes understanding their internal state hard.
- You need semantic zooming to get high-level summaries then zoom in for details to stay in control.
Default To High-Level, Succinct Prompts
- Do default to high-level, succinct prompts and require one question at a time to avoid spam and context rot.
- Keep ground rules short and always present so agents stay focused and scannable.
Use High-Level Tests As Checkpoints
- Write high-level tests (BD D or approval tests) so you can quickly understand what the agent produced at a glance.
- Use approval tests as golden snapshots to accept good outputs and show diffs when things change.
