You pay for two frontier models and route almost everything to one. Both Mac apps are yours for an email, and the guide walks the verbatim prompt, the 65-check list, the job-by-job casting call, and where my own test was unequal.

Substack · Watch/Read

• Nate B. Jones compares Claude Fable 5.1 and GPT-6 Astra, concluding that while one model (Astra) often becomes the primary choice due to its speed enabling more iterative feedback, the other (Fable) remains valuable for surfacing unconsidered questions and aiding in complex thought.
• The speed of a model’s initial output is crucial, as a faster first draft allows for more rounds of corrections and feedback, ultimately leading to a more refined and preferred final product, as demonstrated by the creation of the “Shelf” and “Ledge” Mac apps.
• Nate identifies an “accumulated practice” bias, where a favored model receives more work and context, leading to a learned understanding of the user’s prompting style, potentially making it seem superior to a rival that hasn’t received the same iterative training.
• Users should assign specific roles to different frontier models (e.g., Astra for drafting, Fable for “unsettled arguments”) and focus on the iterative process and the insights a second model can provide, even if its direct output isn’t kept.