If you'd rather watch or listen than read, the video is below.

Jenny Wen, then design lead for Claude at Anthropic, declared that the design process is dead and that we shouldn't trust it. She made the case in a talk in 2025 and returned to the topic in an interview in March 2026, and the line stirred up the design world. NNGroup responded that it's not quite that: the process may be compressed in some situations, especially in the AI era.

I'm a bit late to the discussion, but I'll add my two cents.

For something to die, it has to have been alive. And the implementation of design processes in technology was never a uniform story. It's easier to think in waves than in stable scenarios: each wave left its mark, but the results are uneven. There are companies that use the most refined design strategy available, like Apple and Google, and companies that still build systems the way it was done in the 1990s, having never heard of user experience or usability testing. Many never had a design process to begin with.

So which process died, exactly?

The Design Process Is Dead!

On Lenny Rachitsky's podcast, on March 1, 2026, Wen described what she considers dead: the process taught to designers, in which you do research and discovery, then diverge and converge, and which the field treated as gospel. Her argument is that engineers, with tools like Claude Code, can prototype faster than a designer can sketch. That's why Anthropic works side by side with engineering, in short cycles, and ships unfinished versions to iterate in public.

Up to here, it's an observation about a specific context: an AI company building products from scratch, very fast. The leap comes afterward, when that experience becomes a verdict on "the" design process, as if only one existed.

Waves, Not a Straight Line

The trajectory of design methods in technology is an irregular succession of overlapping layers, not a linear evolution. The design methods movement was born with the 1962 conference at Imperial College, which led to the founding of the Design Research Society in 1966. In Brazil, ESDI, created by decree at the end of 1962, followed the Ulm School model and the industrial design tradition, with no relation to software.

Design entered software development through Human-Computer Interaction (HCI), the field that studies how people use computer systems and designs those systems to be useful, usable, and suited to the people who use them. Three milestones drove this entry: Norman and Draper's book (1986), which popularized user-centered design; Kapor's manifesto (1990), which called the industry's lack of usability its "secret shame"; and the ISO 13407 standard (1999), which defines how to organize the development of interactive systems and presents itself as complementary to existing engineering methods. In Brazil, HCI gained traction in 1998, coupled with the Brazilian Symposium on Software Engineering, and became autonomous in 2000.

The managerial layer came in 2004, with the Double Diamond, created by the Design Council to make the design process understandable to public managers, executives, and technologists. In other words, a pedagogical instrument. In 2008, Tim Brown published his article on Design Thinking in the Harvard Business Review. In Brazil, these ideas arrived through translations, consultancies, and courses: Nubank states that it uses the Double Diamond, and VTEX has a documented case combining the model in design with agile methods in engineering.

Each layer arrived at a different moment, was adopted unevenly across companies and countries, and none fully replaced the one before it. There's no consensus on any of them either. Academic reviews show that the managerial discourse of Design Thinking is superficial and rarely engages with design research, there's disagreement about what it actually delivers, and the most cited evidence of financial return comes from commissioned or correlational studies.

Maturity Lagged Behind the Rhetoric

The available data reinforces the unevenness. In the InVision survey (2019, a self-assessment of 2,200 organizations in 77 countries), 41% were at the lowest level of design maturity and only 5% at the highest. In the McKinsey study (2018), more than 40% of the companies observed didn't talk to users during development. Neither study was peer-reviewed, and the McKinsey one doesn't include software companies. But the numbers converge: most organizations never actually integrated design methods into development.

The "Us Versus Them"

The most persistent point is the friction between design and development. Back in 2006, a retrospective on Kapor's manifesto described organizations where the business defines, programmers build, and designers merely apply an interface on top. The most recent systematic review on the topic, by Zhang and colleagues (2025), confirms that the friction is documented in both directions.

On the development side, designers report being treated as "artists," having suggestions ignored, and seeing the design changed after delivery, without consultation. Developers, in turn, treat user experience as trivial or as the last step of the project. On the design side, there are decisions made without informing the technical team, unworkable proposals, and, in open-source projects, a tendency to criticize more than to contribute. Studies also show designers omitting critical details, ignoring edge cases, or disregarding technical constraints.

Figma's 2025 research shows that AI entered this picture with an asymmetry: 78% say it makes work more efficient, but only 32% trust what it produces, and satisfaction with the tools is 82% among developers and 69% among designers.

In Brazil, the uneven pattern also shows up in AI adoption. According to the TIC Empresas survey, from Cetic.br, usage rose from 13% to 17% of companies between 2024 and 2025, but reaches 50% among large companies and 49% in the information and communication sector. Diffusion is uneven by company size and sector.

In short, in Brazil and worldwide, unequal adoption, whether of design processes or of technologies like AI, is a persistent reality. The success of implementation depends on company size, legal factors, and organizational maturity, among other things. This heterogeneity isn't foreign to design. On the contrary, the field has historically been defined by its capacity to adapt to context.

Which Process of Which Design

Design theorist Christopher Alexander argued that no one becomes a better designer by blindly following a method. What matters, for him, is understanding the idea of creating abstract patterns that resolve systems of forces and combining them freely. Alexander also says that every design problem begins with the effort to achieve fit between two entities: form and context. Form is the part the designer controls, and context is everything that makes demands on it. Good design is a property of the ensemble, not of the form in isolation.

But good fit, he argues, is only accessible through its negative. We can't describe what a house that works looks like, but we can precisely list the misfits: the kitchen that's hard to clean, the lack of privacy, the rain that gets in. Each misfit is a binary variable, it either occurs or it doesn't, and good fit is the state in which all of them are at zero.

In other words: method separated from practice isn't of much use, and the practice of design is messy. Messy, but rigorous. For Mauricio Mejía, a design professor at Arizona State University, design is a series of activities within a process, carried out iteratively and differently each time. That's why there's no single "method" of design. Diagrams like the Double Diamond, chains of blocks, linked hexagons, or arrows convey an order that the real work rarely has. In practice, the path looks more like a tangle of back-and-forth movement. The mess is the visible form of a process that responds to what it discovers along the way, and the rigor lies in deliberate choices, in justifying each return, and in clarity about where one wants to arrive. Linear models work as a communication tool, but taking them as a faithful description of the work creates mistaken expectations about timelines, predictability, and the role of the person designing.

Karri Saarinen, CEO of Linear, turns to Alexander precisely to say that design remains misunderstood because people confuse producing with understanding:

"The hard part of design is rarely generating the form." — Karri Saarinen, Output isn't design

The hard part, according to him, is understanding the problem well enough to know what should exist and how. Context, in this sense, isn't a backdrop: it's the set of forces that defines the problem, such as human needs, technical constraints, conflicting requirements, habits, edge cases, and relationships that only appear to those who take the time to observe them. Bad design arises when those forces are left unresolved. That's why what the design process produces, before anything else, is understanding.

Universal Design and Californian Ideology

It's worth contextualizing where the claim comes from. Anthropic is a Silicon Valley company, and that matters because its premises are loaded with what Barbrook and Cameron called the "Californian ideology": a mix of diffuse ideas, largely drawn from the counterculture movements of the United States (the famous hippies, right?). Its pillars are technology as liberation, individualism, entrepreneurship, libertarianism (distrust of bureaucracy, the state, and regulation), counterculture combined with corporate culture, and faith in the inevitability of technological progress, in which problems that once required political institutions or experts come to be solved by mere innovation.

IDEO, founded in Palo Alto, worked directly with Silicon Valley companies, and the same founder was behind Stanford's d.school. It was in this ecosystem that design, entrepreneurship, and technology came together and helped spread Design Thinking, and it's the same ecosystem that today is the birthplace of Big Tech and frontier AI startups. The Double Diamond is a child of that marriage: a method that presents itself as universal within an ideology that doesn't recognize itself as an ideology.

It's on this ground that the claim makes sense. It doesn't break with the tradition, it radicalizes it. Californian ideology has promised for decades to cut out intermediaries, widen access, and replace human mediation with technical systems. When a design lead at Anthropic says the process is dead, it isn't a contradiction, it's a conclusion. Anthropic itself describes Claude Design as a way for designers to explore more and for everyone else to produce visual work. In principle, the tool doesn't deny design. What it makes dispensable, in practice, is having to go through the designer as an intermediary.

What AI Changes in the Conflict

Up to this point, the conflict between designers and developers was about who should do what: designers complained about being excluded from decisions, and developers, about unworkable proposals or about being called in only to implement decisions made by others. With AI, the boundary between the roles started to dissolve. Developers can generate interfaces, designers can generate code, product managers can prototype. This could reduce the friction and even the destructive dependency between the two fields.

But there's a difference between redistributing capabilities and redistributing power. And part of the problem is that AI produces plausible-looking results in very little time, which creates a sense of completion: it looks like we've arrived at a product, but often the problem wasn't understood, let alone investigated. In the business world this already has a name: AI slop, content produced at scale with little curation, which looks adequate at first glance but tends to be generic, inconsistent, and low quality.

The problems don't stop there. Products reach the market with security flaws, integration problems, and decisions that no one can later explain, because AI tools don't always show how and why they arrived at a given result. They're also non-deterministic: the same request produces different results, which makes it hard to reproduce what worked and turns trial and error into waste. In the end, the form may exist, but the fit between it and a real problem often doesn't.

AI is useful for prototyping, exploring alternatives, and showing possibilities quickly. But that's different from designing. Designing requires understanding the problem, judgment, conversation, confrontation of ideas, effort, and time. The risk is confusing a generated form with a solved problem.

The Impact on the Job Market

The consequences for people who work in design are already showing up in the market. IDEO, a symbol of Design Thinking, cut its headcount by 32% in 2023 and closed its Munich and Tokyo offices, in a reduction that had been underway since 2020 (according to press reports). According to Indeed, UX job postings dropped sharply between the 2022 peak and mid-2023: around 73% in research and 71% in design. In the UXPA survey analyzed by MeasuringU, 35% of organizations lost UX professionals in 2024, double the rate in 2022.

The change is also happening inside organizations. Rosenfeld Media's 2025 survey found a strong relationship between teams that are too small, lack of time to do the work, and burnout. That's why experienced designers report spending more time aligning, persuading, and defending decisions than actually designing.

There's also an age pattern. Employment of software developers aged 22 to 25 in the United States has dropped nearly 20% since the end of 2022 peak, according to the Stanford Digital Economy Lab (preliminary study), and SignalFire measured a 25% drop in hiring of recent graduates at large tech companies in 2024. This data isn't about designers. My hypothesis, and not a measured fact, is that the same logic applies to junior artifact-production work, which is the easiest to automate.

Starting in 2024, the link between AI and layoffs became more explicit. Dropbox cut 20% of its headcount in 2024, in layoffs that hit management, product design, and engineering. Duolingo cut 10% of its contractor workforce at the end of 2023 and published its AI-first memo in 2025. Salesforce cut 4,000 support jobs in 2025, and Atlassian cut 1,600 jobs in 2026. All of them cited AI publicly.

But the records tell a different story. According to Challenger, Gray & Christmas, only 4.5% of the layoffs announced in the United States in 2025 were attributed to AI. And when New York added a field for "technological innovation or automation" to its mass layoff notices, none of the more than 160 notices checked that box, including Amazon's and Goldman Sachs's. One caveat is worth noting: it's a form field, the law hasn't changed, and there may be doubt about what counts as automation. OpenAI's own CEO, Sam Altman, acknowledged in February 2026 that AI washing exists, people blaming AI for layoffs they would have made anyway, though he said he doesn't know the proportion and also expects real job displacement.

The Challenger figure is from 2025. In the 2026 monthly reports, AI has started to carry more weight among the cited reasons, which is worth watching. Either way, citing AI as a reason isn't the same as demonstrating it as a cause.

Conclusion

From this, a few conclusions can be drawn:

  1. There is no single "the" design process whose death can be declared. Talking this way generalizes a specific abstraction, like the Double Diamond, as if it were the entirety of design history.
  2. Maturity lagged behind the rhetoric. Most organizations never actually integrated design methods into software development, and many don't even talk to users.
  3. The friction between design and development is historical and structural. It didn't start with AI.
  4. There's a conceptual misunderstanding. The hard part of design was never giving it form, but understanding the problem to find the fit between form and context.
  5. Attributing layoffs to AI is less solid than it appears. Many cuts publicly attributed to AI reflect macroeconomic adjustments, the correction of pandemic-era hiring, and cost cuts presented as inevitable innovation. Few are formally recorded as automation.

The design process that supposedly died was never alive in any uniform way. The declaration of death, coming from the people building these tools, looks to me less like a technical diagnosis and more like a strategic, market-positioning move.

References