This episode is a practical “try it and measure it” review of GPT-5.2, with lots of quick prompts meant to pressure-test real workflows.
It’s mostly a sequence of demos and comparisons:
- Side-by-side checks vs Gemini 3 Pro.
- Image-heavy tasks (recognition, transformations, and “find the thing” style tests).
- Structured extraction (e.g., table → spreadsheet) and diagram/code generation (e.g., flowchart image → HTML).
- A handful of “app clone” style prompts intended to reveal where the model is strong vs where it still needs careful guidance.