OpenAI initially removed ChatGPT's model picker, angering power users. They fixed this by creating an "auto picker" as the default for most users while allowing advanced users to override it. This is a prime case study in meeting the needs of both novice and expert user segments.
Related Insights
The core technology behind ChatGPT was available to developers for two years via the GPT-3 API. Its explosive adoption wasn't due to a sudden technical leap but to a simple, accessible UI, proving that distribution and user experience can be as disruptive as the underlying invention.
The best UI for an AI tool is a direct function of the underlying model's power. A more capable model unlocks more autonomous 'form factors.' For example, the sudden rise of CLI agents was only possible once models like Claude 3 became capable enough to reliably handle multi-step tasks.
Codex exposes every command and step, giving engineers granular control. Claude Code abstracts away complexity with a simpler UI, guessing user intent more often. This reflects a fundamental design difference: precision for technical users versus ease-of-use for non-technical ones.
Initially, even OpenAI believed a single, ultimate 'model to rule them all' would emerge. This thinking has completely changed to favor a proliferation of specialized models, creating a healthier, less winner-take-all ecosystem where different models serve different needs.
Open-ended prompts overwhelm new users who don't know what's possible. A better approach is to productize AI into specific features. Use familiar UI like sliders and dropdowns to gather user intent, which then constructs a complex prompt behind the scenes, making powerful AI accessible without requiring prompt engineering skills.
Technologists often assume AI's goal is to provide a single, perfect answer. However, human psychology requires comparison to feel confident in a choice, which is why Google's "I'm Feeling Lucky" button is almost never clicked. AI must present curated options, not just one optimized result.
The AMP team believes that accommodating popular user requests like model choice or 'bring your own key' would slow their ability to innovate. They argue that their target users ultimately prefer a superior, opinionated product over peripheral features, even if they ask for them.
Avoid the 'settings screen' trap where endless customization options cater to a vocal minority but create complexity for everyone. Instead, focus on personalization: using behavioral data to intelligently surface the right features to the right users, improving their experience without adding cognitive load for the majority.
An effective Human-in-the-Loop (HITL) system isn't a one-size-fits-all "edit" button. It should be designed as a core differentiator for power users, like a Head of Research who wants deep control, while remaining optional for users like a Product Manager who prioritize speed.
