Hot take: Claude has some of the best AI design I’ve ever seen. The interface is clean, the typography is thoughtful, and the overall “vibe” feels lightyears ahead of the cluttered, utilitarian dashboards we’ve grown used to in the LLM space. But as a designer, I’m constantly reminded of a painful truth: Good product design is more than just a beautiful interface.
True UX is found in the end-to-end experience—the friction (or lack thereof) that occurs when you actually try to get work done. Right now, that experience is hitting two major walls that no amount of elegant CSS can fix.
1. The Token Ceiling
Nothing kills a creative flow faster than hitting a usage limit mid-sprint.Imagine you’re deep in the “zone,” working through a complex architectural solve, and the tool suddenly goes dark. When the “brain” of the product shuts off just as you’re reaching the finish line, the utility of that beautiful UI drops to zero.In a professional environment, reliability is the ultimate UX. If a tool can’t guarantee it will be there for the duration of a task, the polish starts to feel like a distraction from the instability.
2. The “Black Box” Problem
There is a growing frustration among power users regarding the lack of manual tweaking. High-level automation is a magic trick the first ten times you see it, but as professionals, we eventually need granular control. We need to be able to:
- Turn the knobs on specific parameters.
- Adjust outputs manually without starting from scratch.
- Fine-tune the logic rather than just “shouting into the prompt box” and hoping for a better roll of the dice.
When we lose user agency, the tool stops being a collaborator and starts being a gatekeeper.
The Takeaway
A product can look like the future, but if the operational constraints (tokens) and the lack of user agency (manual control) stand in the way of the work, the “experience” isn’t quite there yet.
Beautiful UI gets users through the door, but reliability and control are what keep them in the room.
If AI tools want to move from “fun experiments” to “essential infrastructure,” they need to prioritize the invisible parts of design: the limits, the levers, and the long-term flow.
Leave a comment