The seduction of speed is making bad UX acceptable
We need to stop this.
Teams are moving faster than ever. AI can generate interfaces in minutes that look polished and complete.
That speed feels like progress. It feels like clarity, like we are closer to the answer.
But we are often not.
What is actually happening is simpler and more dangerous.
Output that looks finished is being accepted as finished. Quality is being judged by fidelity instead of function.
The “Done Illusion”
AI-generated interfaces create a false sense of completion.
A rough sketch invites critique. A polished UI shuts it down. When something looks real, teams hesitate to question it.
That is how bad decisions get locked in.
This is not just intuition. It is psychology. When something is easy to process, our brains are more likely to assume it is correct, a concept known as cognitive fluency (Reber, Schwarz & Winkielman, 2004).
Under the surface, much of this output is weak.
Unless your design system is explicitly built to work with agents, these tools rely on limited components and familiar layouts. They cannot reliably identify which patterns are proven, repeatable, and appropriate for a given product or industry.
They also fail at consistency.
The same task can show up in multiple places across a product and behave differently each time. The system does not recognize that these are connected moments. It treats them as separate problems.
This creates real damage.
Cohesion breaks.
User mental models are ignored.
Muscle memory disappears.
Users depend on consistency to move quickly and confidently. When every interaction is slightly different, the product becomes harder to use.
Vibe coding is not design
Vibe coding is useful. Product managers should use it.
Being able to quickly generate something realistic enough to react to improves collaboration. It makes ideas tangible. It speeds up conversations.
This is not new.
We saw the same shift when PMs started making wireframes. Visual thinking helped teams align faster. The problem was never the wireframe.
The problem was treating the wireframe as the solution.
Vibe coding is the same pattern with higher fidelity.
It is thinking out loud with a rendering engine.
That is valuable.
But it is not design.
The shortcut that breaks design
The real risk is the shortcut:
From “I had an idea” to “we have a design”.
That jump skips the part that matters most. The messy middle where real design work happens.
That middle is where teams explore multiple approaches, challenge assumptions, test patterns against real user behavior, make deliberate trade-offs.
Design is not the output. It is the process that gets you to the right output.
That process requires skill. It requires training. It requires time to generate and evaluate options before committing.
Most people are not trained in this. Even very capable PMs.
Without that process, teams default to the first thing that looks good.
What’s missing? Checks and balances
We need to put real checks and balances back into the system.
Right now, vibe coding skips them.
We need mechanisms that ensure we’re solving the right problems in the right way. Even when a problem is valid, the experience still has to guide users through it clearly, predictably, and efficiently.
That means asking harder questions before we ship:
Does this solve a meaningful user problem?
Is the workflow easy to understand without explanation?
Does it match how users expect this task to work?
Are we using the same patterns for the same types of tasks across the product?
Where can users make mistakes and how are we preventing or recovering from them?
Consistency is not optional. If a task repeats, the pattern should repeat. If a behavior is learned once, it should apply everywhere.
This is how we reduce cognitive load. This is how we build muscle memory. This is how we reduce errors and mitigate risk.
Right now, vibe coding does not enforce any of this.
There is no built-in validation of user mental models. There is no guarantee of pattern consistency. There is no systematic check for error states or edge cases.
The output looks complete, but the safeguards are missing.
That is the gap.
Why this matters more now
High-fidelity output changes behavior.
When something looks finished, it is harder to question.
Teams converge too early. They stop exploring. They commit before they understand the problem fully.
Speed amplifies this.
It does not just accelerate progress. It accelerates decisions. Good or bad.
Right now, it is accelerating the bad ones.
What vibe coding is actually for
Vibe coding should be treated as raw material.
It is a way to externalize thinking quickly. It is a starting point for discussion. It is not the answer.
Designers should welcome PMs who bring these artifacts forward. It makes collaboration stronger.
PMs should expect those artifacts to be challenged, reshaped, and improved through design.
This is not a turf issue.
It is how teams work together to build better products.
Human judgment still matters
Some things have not changed. Users still rely on consistent patterns. Familiar structures still build trust and efficiency. Good design still requires exploration and judgment.
AI can generate interfaces.
It cannot decide which ones should exist, why they should exist, or how they should behave as part of a system.
That responsibility is still human.
That responsibility is still design.
Speed is not the problem.
The problem is accepting poor design because it looks finished.
We need to stop this.