David Ondrej describes his experience using Claude Fable 5.2, including usage limits, reusable prompts and the trade-off between reasoning effort, speed and cost. His comparisons reflect his own workload and reported benchmarksEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria. rather than a controlled general evaluation.
Ondrej explains an experiment in using a more capable model to clarify a difficult task before switching to a faster model for execution. He also describes using Gemini to review changes made by another modelCode review examines proposed source changes to assess their correctness, clarity and suitability before or around integration., while presenting the benefits as his own experience.
The discussion distinguishes cost per token from cost per completed taskCost per completed task measures the total AI, tool, infrastructure, retry, and repair expense for each verified useful outcome. and considers agent loops that run checks only when needed. Model rankings, pricing judgments and predictions are the presenter's views at the time of recording.
Watch on YouTube
