Performed Identity vs. Embedded Identity
OpenAI recently introduced a Personalized Voice feature — the ability to shape how ChatGPT speaks to you. It's a compelling product decision. But notice what it implies: identity is a setting, something you configure from the outside.
Anthropic, by contrast, makes no such offer. Their stated value is human-centered AI — and that philosophy isn't a feature you turn on. It's structural. It's in how Claude reasons, defers, and holds its ground. You don't configure Claude's character. You encounter it.
OpenAI / ChatGPT
Identity as a feature — personalized, configurable, performed on request. The brand signals capability. It becomes what you need it to be.
Anthropic / Claude
Identity as architecture — consistent, present, not a mode you switch on. The brand signals character. It stays itself regardless of the prompt.
This is what I call identity drift — when a brand chases features without a coherent self, it begins to shape-shift. Users feel the absence even if they can't name it. Structurally strong brands don't need a personality slider, because the personality was never missing.
The brands with the strongest identities are structurally sound first — expression follows architecture, not the other way around.