Anthropic
Strong friction and qualification. Lived stake is where the score drops.
After I published the four-component framework, people asked for a real example. I ran Anthropic through it first because the company is instructive. All four components are present, at very different levels of authenticity.
Anthropic scores 7.5 out of 10.
Where the story is strong
The friction is clear and shared. AI capability is racing ahead of anyone's ability to keep it safe, and developers, enterprises and regulators all feel it. Anthropic gets specific: models are black boxes, and safety tends to get bolted on after training.
Constitutional AI earns the 8.5 for qualification. The structural insight is that human feedback alone teaches models to please raters, while an explicit constitution makes alignment something you can inspect and scale.
Where it falls short
Lived stake is the weak point. The origin story is a departure from OpenAI over safety priorities. Set it beside Brian Chesky, who couldn't afford rent and wondered whether others had spare rooms to monetize, and Anthropic's version reads as concern about where the industry was heading. That's an observer's position. The story would score higher if it showed specific alignment failures the team hit and the experiments that exposed the limits of existing methods.
The aspiration is clear but hard to join. When Stripe told developers they shouldn't have to become payments experts, every developer knew the next move. If you ask what Anthropic's narrative asks enterprises or regulators to do today beyond choosing Claude, the answer is thin.
The emotional cost is missing too. The writing is professional and measured, with little of the frustration that comes from building systems you can't fully control.
What would raise the score
Tell concrete origin stories from inside the work. Publish tools and evaluation methods other teams can use. Spell out what developers, enterprises and regulators should do next.
Anthropic's narrative is coherent and differentiated, and it hasn't yet made the right behavior feel inevitable. That's the distance between 7.5 and 9.5.
Want your own score? A Narrative Diagnostic runs your website, messaging and leadership content through the same four components.
