Summary: UX shifts radically as "vibe coding" places users in control of digital functionality. AI enables rapid, inexpensive development, empowering domain specialists to design software independently. UX professionals must pivot to scaling effective user innovations, guiding emergent patterns, and embracing a strategic role instead of a design role.
The new trend toward vibe coding and vibe design may be upending the user-centered design paradigm that has remained based on the same ideology since the first UX design projects at Bell Labs, starting in 1947. When AI hands keys to everyday users, the priesthood of professional designers topples amusingly into irrelevance. Embrace vibe-driven chaos, but audit your UX debt meticulously. In a future flooded with quirky, spontaneously designed software, usability remains vital — lest vibe drowns the users.
If you don’t have time to read this, admittedly long, article, watch my 3-minute video summary on YouTube.
I should start by admitting to a major mea culpa: In one of my first articles about AI in May 2023, I correctly identified that AI is a new user interface paradigm based on intent-based outcome specification instead of interactive commands. I didn’t realize that the consequence of taking this insight to its logical conclusion: it’s not just the UI that changes, the very nature of digital products changes, leading in inevitable changes in the product design lifecycle.
The simple part of intent-based outcome specification (that I got right two years ago) works within the framework of an existing application and simply changes its user interface. For example, instead of using color pickers, drawing tools, and other commands to make a cartoon about user-driven design, I can tell Ideogram “cartoon showing a robot working at a computer. Behind the robot is a businesswoman in a suit saying "I need a dashboard to track shipments across regions".” Ideogram figures out how to fulfill my intent. I never had to issue commands to pick the colors, but if I don’t like its choices, I can iterate on the prompt to clarify those details of my intent.
Vibe design uplevels intent-based outcome specification to allowing users to specify the functionality they need, after which the AI designs an appropriate UI for those features. (Ideogram)
The next level of intent-based outcome specification (which I overlooked until now) is that users drive the design of the functionality itself, not just their use within an application designed by others. Vibe coding and vibe design empower domain specialists to make computers do their bidding and create the applications they need for their work, hobbies, or lifestyle.
I should give a very large hat tip to Georg Zoeller (VP of Technology for NOVI Health in Singapore) for alerting me to this newer and larger scope of user-driven design.
Old Assumption: Design and Development Were Expensive and Slow
Until the advent of AI-supported (or AI-delivered) software development, we relied on scarce supplies of highly paid human programmers to implement anything. We relied on even more scarce supplies of equally expensive interaction designers, user researchers, and product managers to specify what those developers should build.
The economy always aims at optimizing the use of scarce and expensive resources. In our case, this meant that relatively little software was built and that minimizing the amount of code change was imperative. In particular, all aspects of the user experience, from the feature selection to the user interface to those features, needed to be specified in advance. Since designers and researchers were also expensive, only limited design iteration was possible, even though we’ve always known that extensive use of iterative design was the only way to create high-quality user experience.
Most of all, it was incredibly expensive to make changes to an application after launch: most software engineering experts estimated the cost of a design change to be 100x higher if made to a shipping product than if made during the prototyping stage before any code was written or final designs created.
In 1989, I developed Discount Usability as my response to these economic restrictions: If we limited each round of user testing to 5 study participants and created fairly limited prototypes of each design iteration, a product team with a handful of UX professionals on staff might be able to make it through three or four — or maybe even five — design iterations before having to deliver design specs to development. Heuristic evaluation further expedited this process because some design ideas could be evaluated relative to the 10 heuristics, saving a round of user testing. My push for simplified usability methods allowed more project staff to contribute to UX design, saving on those scarce UX professionals.
All very good, and I still recommend this approach for legacy companies that have not pivoted to going AI First.
New Assumption: AI Creates Abundance and Speed
Current AI tools for software developers often achieve an acceleration of 4x to 10x, and leading startups now have 95% of their codebase generated by AI. (That 5% human-created code is essential, though, because it makes the rest work and forms the base for a solid software architecture.)
But as the First Law of AI says, today’s AI is the worst we’ll have in the rest of our lives. When AGI is achieved around 2027 (and especially when superintelligence arrives around 2030), AI is expected to accelerate software development even further, maybe by up to 100x.
This acceleration has three implications for user experience:
When software development becomes 100x faster, we can afford to experiment more, and it doesn’t become as essential to compress development cycles to cur corners on design because they’ll already be plenty fast.
When software becomes 100x cheaper, more will be built, leading to more demand for good design. That’s why I’m not worried about unemployment despite everything else I say in this article.
Software development becomes much easier, when AI takes over all of the heavy lifting. This makes it possible for users to do much of their own design and development, using that infamous vibe coding and vibe design.
Instead of waiting for a priesthood of professional software developers and UX designers and researchers to slowly create software for their needs, users can make their own applications. As a result, we’ll have immensely more software in the future than we ever had in the past, and we must evolve new UX methods to accommodate this software explosion.
The software explosion: AI will create millions of times more software than we ever had, just as the Internet created millions of times more video than was ever made by Hollywood studios. (Leonardo)
Software As Content: The Power Law for Audience and Quality
I previously wrote an article about service as software: when AI achieves higher intelligence and more knowledge than professional consultants, many services will be performed by software instead of humans. That means that many small companies and individual consumers will get the high-level advice that was previously reserved for high-paying enterprise clients.
But software itself is also poised for a mass-market revolution, as it becomes dirt cheap to design and develop. Content forms a good analogy: Before the web revolution, books and newspapers were published by a small number of companies, meaning that very few authors saw print. Filmed entertainment (and educational video) was only produced by a handful of big studios, concentrated in locations like Hollywood and Bombay.
Now, anybody can publish a newsletter like the one you’re reading, and a worldwide readership is reached at the press of a button. (I have readers in 148 countries.) Same for video: it takes a few hours to film a video (or generate with AI video tools), edit, and upload to services like YouTube, TikTok, and the like, where any creator can reach a large audience if the videos are entertaining or useful enough.
Internet publishing broke the content oligopoly.
The same will be true for software, once everybody can create applications without the need for professional software developers or professional UX staff. I can already hear my readers say, “yes, but those apps will be crap, if made without professional help.” True enough, according to Sturgeon’s Law, which says that “ninety percent of everything is crap.” That’s true for blogs and YouTube videos as well, and yet YouTube videos are watched for approximately one billion hours every day. There are more than half a million hours of new video uploaded to YouTubve every day, and even if 99% are crap (a “super-Sturgeon’s Law” we might hypothesize for Internet content), that would still mean 5,000 hours of good new videos for you to watch every day.
Not only is there a “long tail” type of power law for the size of the audience consuming each content piece, there’s probably a similar distribution for the quality: small amounts of great stuff and larger amounts of progressively worse content. However, even slightly bad content still finds an audience if it addresses a niche need or interest.
Same with software: an app made by a user for his or her own use might never see another user, and so will have an audience of one. Clearly not feasible to build as an old-school product development project. But if that user “vibes” it into existence and if it a present and urgent need in the moment, that’s enough.
Some user created software will have wider applicability. Maybe be used by 10 or 100 users within a department or among the dentists in a certain town. The best of that town’s dentist-created software will be demoed at the next dental convention and spread to other towns, gaining thousands of users.
Vibe design: just speak your needs and the AI makes your desires into code and designs a working application for you. (Midjourney)
User-Driven Design and UX
User-driven design may seem like it solves the biggest problem in UX design: how to discover and solve user pain points. The users know their problem and if they come up with a solution that works for them, then we’re done. No need for any UXers to scratch their brains or conduct expensive user research.
However, things are not that simple. You may know that one of my main UX slogans is “You ≠ User.”
You are not the user. Different mental models, different computer knowledge, different skills, different everything. (Leonardo)
Even though I’ve used this slogan for the last few decades, it is a simplification of my original slogans from my book Usability Engineering, published in 1993:
Designers are not users.
Users are not designers.
Point 1 is why we do user testing: it’s not enough that the designer thinks that the UI is easy. We must check with the actual users.
I don’t talk about Point 2 nearly as much, other than to remind readers that this is why we have to do studies where we observe users, as opposed to simply asking them how much they like various design mockups and ship whatever version is liked the most.
However, the deeper insight in Point 2 is that it used to require special UX expertise to deconstruct user pain points and architect a coherent design to solve the problems. Since users aren’t designers, we can’t expect them to be able to do this.
In the past, these points were true, and they would still be true if we relied on asking a single user to design the software (say, we got hold of a dentist and asked him or her to design the software solution to run a dental practice).
Users are experts at being users, but not at designing something that solves their problems, much as they feel those problems everyday.
One approach to bridge the gap between designers and users was called participatory design. The idea was to invite representative users to participate on the design team and represent user needs in every design meeting. Unfortunately, this didn’t work. When I worked at the telephone company, we had so-called SMEs (subject-matter experts) who were people who used to climb telephone poles but had now been at Bell Labs for years. No surprise, each SME quickly became indoctrinated in the software development way of thinking and was no longer an average telephone repair tech.
I expect user-driven design to turn all these problems around, though it may present us with new ones.
Vibe design will certainly not turn users into skilled UX designers or accomplished software engineers capable of architecting an enterprise solution. But they’ll solve their own problems, and the best of these solutions will then evolve into more widely-used tools. That’s where the pros will come into play, when rearchitecting the amateur solutions for scalability.
Also, the vibe designs won’t be as bad as the amateur design of the past. Users will be using AI tools to create their solutions, and these tools will embody the 50 years of usability best practices the UX field has accumulated. In particular, visual design and UI design will likely be better in a few years when done by AI than when done by human designers, because no human can remember all of the tens of thousands of usability guidelines or visual design rules and contextually apply them correctly to the problem at hand. But a next-generation AI with a huge knowledge base and a good amount of reasoning compute will solve all those design problems correctly.
No human can remember all the usability knowledge accumulated over the last 50 years of UX research in sufficient detail to apply it correctly for any context. (Midjourney)
Visual design, from icons to layouts, and UI designs, from menus to workflow, will mainly be done by AI trained on the field’s cumulated usability and design knowledge. This means that “amateur” user-driven design will be better than most of the UIs made in the past by professional UX designers. (Leonardo)
Feature design will also improve, for two reasons: first, user-driven design is inherently based on the features each user needs, and second, those solutions that spread also have features that other people need.
Likely Changes to UX
I still believe in the need for UX professionals to scale the best of the user-driven designs, but the UX profession must change in the following nine ways:
1. Usability Shifts from Gatekeeping to Empowerment: In the past, usability professionals acted as guardians of quality, compensating for slow, costly development cycles with rigorous processes. Today, as business experts rapidly prototype, the role of usability specialists transitions to enablers. They must embed core principles — consistency, error prevention, feedback — into the AI no-code tools, templates, and AI-assisted design systems. This will help user-driven design become better out of the gate.
2. Discount Methods Become Even Leaner: My original discount usability engineering (heuristic evaluation, 5-user testing, iterative design) was designed for efficiency. These methods now demand further streamlining. Think micro-testing: 1–2 users providing feedback in real-time during prototyping sessions, or AI analyzing clickstreams to flag usability risks. Usability professionals will curate these tools, ensuring they deliver actionable insights without interrupting rapid cycles.
Microtesting with a single user allows us to put even the smallest design issue under the microscope. (Ideogram)
3. From Prevention to Correction: The economic consequences of bad design have fundamentally changed. When fixing issues took weeks of developer time, preventing design errors is paramount. When fixes take seconds, the balance shifts to: (a) Less emphasis on getting everything right before release, (b) More focus on quick identification and correction, and (c) Accelerated learning from real user behavior rather than anticipated problems. This doesn't mean abandoning preventive measures, but recognizing that perfect prevention is less critical when correction is nearly free.
4. User Research Focuses on Bridging Gaps: Business professionals know their personal workflows but may overlook broader user needs. Usability experts must conduct targeted research to uncover blind spots: Are assumptions about “what users need” validated across job roles and countries? Rapid, automated surveys or sentiment analysis of user-generated feedback can surface mismatches between designer intent and real-world use.
5. Combatting Usability Debt: Speed risks accumulating usability debt—minor flaws that compound into systemic frustration. UX experts will act as auditors, identifying recurring issues (e.g., inconsistent navigation in 30% of prototypes) and refining the AI tools to preempt them.
6. Focus on UX speed: The advantage of vibe coding is speed. Discount usability in this context means embracing continuous, micro-testing. Instead of dedicated rounds of user testing, usability feedback should be integrated into every iteration, however small. This aligns with the “build fast, fail fast” mentality, but with a crucial caveat: learn fast. When releases happen continuously, UX work must do the same.
7. Core Task Usability First, Vibe Polish Later: In a rush to market with vibe-driven products, there will be immense pressure to quickly release something. Usability efforts must prioritize ensuring that the core tasks users need to accomplish are actually possible and reasonably efficient, even if the vibe or polish is still lacking.
8. Minimum Viable Usability: Define the absolute minimum level of usability required for users to achieve their essential goals. Focus testing and iteration on reaching this “minimum viable usability” threshold first. Cosmetics and advanced features can come later.
9. Prioritize Functionality over “Fun-ctionality”: While vibe design may imply a focus on engaging and delightful experiences, remember that users ultimately want to get things done. Professional UX efforts should guard against prioritizing fun at the expense of core functionality and ease of use. A product can have great vibes but be utterly useless if users can’t figure out how to operate it.
UX work is about to experience 9 dramatic changes. If you’re a design leader, now is a good time to take your team on an offsite to prepare them for change, because once it happens, it’ll be too late. (Midjourney)
Speculative Changes to UX
We must consider the extreme possibilities if vibe coding and vibe design truly become the main way to create software for both business applications and personal use. Here are some ideas that I do not recommend for now, but which describe possible future scenarios that may happen in 20 years. (Remember that 20 years is only half-way through the 40-year career of a UX newbie who started this year, so it’s absolutely relevant to discuss such scenarios even if they seem far-fetched now.)
AI-Powered “Vibe Analysis”: Imagine algorithms trained on vast datasets of user reactions to designs, attempting to predict the “vibe” of a new interface. The UX professional’s role would then be to interpret and critique the AI’s output, ensuring it doesn’t lead to homogenous, soulless design.
Intention-Based Design: Vibe design may be a temporary phase, and the long-term future is one where nobody ever directly designs anything. Instead: (a) Users express business needs and the system designs an appropriate solution, (b) The interface learns from user frustration and adapts without explicit changes, and (c) UX staff focus on the meta-systems that translate intent to design and not on the products themselves.
Crowdsourced, Continuous Usability Feedback in Vibe Communities: If vibe design becomes community-driven and development is incredibly fast, perhaps usability evaluation will also become continuous and crowdsourced. Imagine “vibe communities” constantly providing feedback on prototypes, not just on aesthetics, but also on ease of use. This could be a very noisy signal, but aggregated and analyzed by experienced UX professionals, it could provide real-time usability insights at scale. Think of it as a massively distributed, always-on hallway test.
Crowdsourced, continuous usability feedback may feed directly into an AI designer and cause a living user interface to immediately adjust to user needs. (Leonardo)
Usability-as-a-Service API: Vibe coders could plug their designs into an API that provides instant, automated usability feedback based on heuristics and pre-defined user profiles. This is essentially a more sophisticated version of the embedded usability principles mentioned above.
The Rise of the “Design Influencer”: In a world obsessed with aesthetics and quick takes, perhaps usability expertise will be disseminated through short-form videos and social media posts. The “Design Influencer” will offer bite-sized tips and tricks, catering to the attention span (or lack thereof) of the vibe designer.
The landgrab may soon be on for becoming one the design influences of the future. (Leonardo)
Emergent Usability: A truly radical concept is that usability will somehow emerge organically from the collective vibes of the users and creators. This is a form of design by pure intuition and luck, and it’s a truly radical future that fully abandons creationism in favor of relying on evolution. Evolution worked for nature, albeit with a detour around the dinosaurs, and so maybe great software can come into being along similar lines: it’ll emerge from the mud without the need for intelligent design(ers).
Who knows, maybe a great design will emerge from the evolutionary muck of millions of user interfaces. (Midjourney)
Six possible radical changes to UX may change over the next twenty years.
Instead of the “emergent usability” scenario, I think a more plausible future is that Sturgeon’s Law will remain in force and 90% of the vibe designs will remain bad. UX professionals will have the job of distilling all that crud and sift off the 9% good design and the 1% great design that did come out of user-driven design.
Exploring the expanding universe of UX brought into creation by vibe design. We’ll no doubt discover strange new worlds and new civilizations of UX methods that never made for Earth-bound inhabitants of the old scarcity-controlled UX profession. However, unless UX professionals change at warp speed, they’ll be left behind. (Midjourney)
10-Year Prediction: Expanded Strategic Influence for UX
The 20-year view I just presented wasn’t a prediction, but a set of possible scenarios. Limiting myself to a 10-year horizon, I think the following is a credible prediction for the UX profession:
With mundane design tasks delegated to AI, UX professionals can aim higher in the value chain. They can devote more energy to research, strategy, and system-level design. For instance, defining the overall design system and guiding principles that the AI should adhere to, or focusing on complex experience problems (like emotional design, trust, and ethics) that require human insight.
UX staff might also take on a product leadership role, since understanding users becomes a key differentiator when AI can generate the basics. In organizations, UX experts could increasingly act as the bridge between what users need and what the AI builds — a role that’s both strategic and creative.
Even after AI takes over all traditional design work, I suspect there will still be a role for human UX professionals in keeping a keen eye on its efforts and defining the strategy for future work. (Leonardo)
Watch the 3-minute video summary of this article (YouTube).