The Daily AI Show Using ChatGPT 5.2? Better watch this first!
30 snips
Dec 12, 2025 The discussion kicks off with Brian's deep dive into GPT 5.2, revealing how it enhances goal-driven prompting and simplifies building custom models. They celebrate Topaz Labs' Emmy for AI restoration and explore Science Corp's groundbreaking retinal implant funding. Improvements in spreadsheet reasoning and chart reading showcase significant potential for professional analysis. A standout moment features Brian demonstrating innovative phase-based prompting, making complex builds more accessible, while reflections on the future of AI hint at exciting developments ahead.
AI Snips
Chapters
Transcript
Episode notes
High-Level Goal Prompting Works Better
- GPT 5.2 shifts models toward high-level goal-driven prompting, reducing the need for brittle router/master prompts.
- Brian rebuilt his internal gem builder in hours using far less prompt scaffolding to test this change.
Weigh Latency Versus Depth
- Consider latency and timeout tradeoffs when using 5.2 Pro for long reasoning tasks.
- Use it when deep, expensive reasoning is worth the time and cost, otherwise choose faster models for quick answers.
Read The 5.2 Prompting Cookbook
- Read the GPT 5.2 prompting cookbook before building to understand behavioral changes and best practices.
- Use the cookbook's patterns (clarification prompts, internal outlines, anchor claims) to reduce hallucinations.