Even as AI models become vastly more powerful, widespread adoption is throttled by the slow evolution of users' mental models of what AI can do. People rely on a system based on past experiences, and it takes a 'magical' result to expand their belief in its capabilities for new, complex tasks.

While public discourse on AI models often focuses on incremental improvements in common tasks like writing emails, the most profound advancements are happening in specialized fields like science and mathematics. This capability gap creates a disconnect in perceived progress.

A major focus for OpenAI's design team is the growing gap between what their models are capable of and what users actually know they can do. The design team's job is to create interfaces and tools that expose the model's full potential to the user.

Despite models demonstrating PhD-level capabilities, most people only use them for basic tasks. The biggest hurdle for AI companies is not making models smarter, but bridging this usability gap by making advanced power easily accessible to the average person, likely through better interfaces and agents.

While GPT-5.5 is a massive technical improvement, it may not feel transformative for 99% of users' daily workflows. Previous models like GPT-5.4 were already proficient enough for common tasks. The new model's value is realized at the ceiling of capability, on complex edge-case problems that stressed older models, rather than in everyday use.

The main barrier to AI's impact is not its technical flaws but the fact that most organizations don't understand what it can actually do. Advanced features like 'deep research' and reasoning models remain unused by over 95% of professionals, leaving immense potential and competitive advantage untapped.

Fable 5's extreme thoroughness, while powerful, makes it unsuitable for tasks like writing product specs. Its outputs are too dense and detailed, missing the bigger picture in a way that can delay shipping. Sometimes a "dumber," more pragmatic approach is more effective for product development.

Anthropic's research shows that experienced AI users get more value because they learn to interact with the model as a collaborator. Proficiency is not just prompt engineering, but a learned skill of engaging the AI in a more sophisticated, iterative partnership to explore ideas.

Since current AI is imperfect, building for novices is risky because they get stuck when the tool fails. The strategic sweet spot is building for experts who can use AI as a powerful but flawed assistant, correcting its mistakes and leveraging its strengths to achieve their goals.