Summary
Nate B Jones of AI News & Strategy Daily compares ChatGPT 5.6 “Soul” with Anthropic’s Fable 5 — which he calls part of the “Mythos” model lineage — arguing that benchmark scores are an increasingly poor guide to model selection. He references his private “Dingo” benchmark suite, on which Soul scored 93, and notes Soul topped the AgentSlash exam across 55 professional fields. Despite those results, Jones finds himself reaching for Fable 5 when tasks require ambiguity-handling and conceptual depth, which he attributes to Anthropic’s emphasis on large-scale pretraining.
The video’s central thesis is that AI models are evolving into distinct “families” rather than a linear intelligence hierarchy. The OpenAI 5.x family — including Soul, Terra, and the cheaper Luna series — excels at explicit, long-running agentic coding workflows with precise instruction-following. Anthropic’s Mythos/Fable lineage, by contrast, is described as more philosophical, front-end-savvy, and adept at reading between the lines. Jones also references Anthropic’s published “J-space” research, which examines how its models computationally manipulate higher-order concepts during token prediction.
The practical takeaway Jones offers: don’t start by picking a model — start by examining your own best work and the process that produces it, then find the model that accelerates that loop. Long, detailed voice-prompted workflows favor Soul; high-level ambiguous briefs favor Fable; cost-sensitive coding tasks may favor Luna. The video is aimed at knowledge workers and professionals evaluating subscriptions across multiple frontier model providers.
📺 Source: AI News & Strategy Daily | Nate B Jones · Published July 13, 2026
🏷️ Format: Opinion Editorial
