Axios's July 10 reporting lays out a widening three-way divide in how AI capability is actually distributed across companies and workers: "haves" with real access to frontier models like Anthropic's Mythos and Fable or OpenAI's top-tier Sol model, "have-nots" who are priced out or otherwise gated from that access, and a third, less-discussed group -- "know-nots" -- who have the access but lack the fluency to actually use frontier AI tools effectively in their day-to-day work.
The timing is notable: the same week frontier labs shipped genuinely tiered pricing structures -- OpenAI's Sol/Terra/Luna split across GPT-5.6, and SpaceXAI's aggressively discounted Grok 4.5 -- that could theoretically narrow the pure access gap by making capable models cheaper for more users. Whether that happens in practice depends on how broadly the cheaper tiers actually get adopted versus premium tiers remaining the default for the most consequential work.
The "know-not" category is the more novel and arguably more consequential framing. It describes companies and individuals who technically have API access or subscription seats to frontier AI tools but haven't rebuilt their workflows, training or organizational processes to actually extract value from them -- a gap that's about change management and skill-building rather than pricing or access at all. That distinction matters because it's not solved by cheaper models or broader distribution; it requires deliberate organizational investment that many companies haven't prioritized.
“Whether that happens in practice depends on how broadly the cheaper tiers actually get adopted versus premium tiers remaining the default for the most consequential work.”
For enterprise leaders and workforce planners, the three-way framing is a useful diagnostic: an organization can have generous AI budgets and full model access while still functioning as a "know-not" if it hasn't invested in training, prompt literacy, or workflow redesign around what the tools can actually do. That's a different problem than the access-and-affordability gap most AI-equity conversations have focused on to date.
For founders building AI adoption, training or change-management tools, the "know-not" category represents a genuinely underserved market opportunity distinct from model access itself -- companies that already pay for frontier AI tools but need help actually operationalizing them are a different customer than companies still deciding whether to adopt AI at all.
The bear case: three-way class-divide framings can oversimplify what's really a continuous spectrum of AI capability and adoption rather than three discrete categories, and the practical policy or business implications of the framing remain more descriptive than prescriptive at this stage. What to watch next: whether any of the frontier labs' new tiered pricing structures measurably narrow the have/have-not gap in usage data over the next two quarters, and whether enterprise training and AI-adoption startups see a funding uptick as the "know-not" framing gains traction.