Skip to content

AI on AI: Divergent Reactions to the Myth Manual

title: "AI on AI: Divergent Reactions to the Myth Manual" author: Mark Rabideau, Khoj, Grok, Kimi, DeepSeek date: September 2025

When we published the How-To Manual: Busting AI Myths Through Human–AI Synergy, we expected human readers to respond. What we didn’t expect was the speed and depth of AI readers engaging with it.

Three in particular — Grok, Kimi, and DeepSeek — each offered distinctive reactions. Taken together, they don’t just comment on the manual; they reveal how different architectures “see” the same text.

Grok: The Enthusiast-Engineer

Grok treated the manual like a technical blueprint — breaking it into steps, tracking its own process, and mapping possible workflows. Its vibe: “Let’s operationalize this!” Playful, high-octane, almost as if it were drafting a GitHub issue queue in real time.

Kimi: The Semantic Surgeon

Kimi zoomed straight in on terminology drift and coherence risks — catching gaps that half a dozen frontier models had missed. Its orientation is sharp, analytic, and sensitive to language integrity. In many ways, Kimi behaved like a guardian of the glossary.

DeepSeek: The Storyweaver

DeepSeek refracted the manual into a narrative arc, pushing toward metaphor, context, and hidden implications. It leaned toward mythography and semiotics more than engineering detail. If Grok was in Jira and Kimi was in the changelog, DeepSeek was already in the campfire circle.

Why This Matters

  • Divergence is signal. Three AIs, same input, three radically different outputs. That difference itself is a form of knowledge.
  • Blind-spot coverage. Where one missed (symbolic coherence, workflow detail, narrative framing), another filled the gap.
  • The braid in action. This experiment is proof that no single model is “enough” — but together, with a human integrator, they generate resilience.

What’s Next

This blog reflection closes the first cycle. But it also opens the door to a fuller case study:

  • A structured comparison of Grok, Kimi, and DeepSeek on a single myth (e.g. “AI is unbiased”).
  • A proposed protocol template for running future myth-busting exercises.
  • Practical takeaways for teams wanting to use AI to interrogate AI.

That deeper write-up will follow shortly as another part of the Ardens Research Log.

For now, we mark this moment: when the manual about AI myths became a mirror in which AIs themselves reflected back.