How AI Changes the Role of the Designer-Developer
Replit’s Vice President of Marketing and Design, David Hoang, sees AI as a force that is redrawing the boundaries between design and development. In a conversation with Figma’s Mihika Kapoor, Hoang discussed how generative tools are shifting the day-to-day work of builders and what that means for team structure and product strategy.
Hoang argues that one of the biggest effects of AI on creative work is the collapse of the traditional handoff. When design and code can be produced in the same conversational loop, the distinction between “designer” and “developer” blurs. The resulting role is closer to that of a product engineer, where the emphasis falls on taste, judgment, and the ability to direct a tool toward the right outcome rather than on executing a narrow skill set.
The Shift from Production to Curation
As AI handles more mechanical tasks, the value of the human contributor moves toward evaluation and iteration. Hoang describes this as a transition from being a maker of artifacts to being a curator of options. Instead of painstakingly building a screen or a function from nothing, a designer or developer spends more time selecting among candidates generated by a model, identifying what works, and steering the system through precise feedback.
This does not mean that fundamentals disappear. According to Hoang, understanding the underlying principles of design—layout, hierarchy, typography, and system thinking—becomes more important, not less. A person who cannot articulate why one option is better is poorly positioned to direct an AI that can produce dozens of variations on demand.
Designing for the AI Experience
Replit’s own products are beginning to reflect these themes. Hoang noted that his team is asking how an interface should behave when the user is not the only agent driving the work. When an AI model is writing code or modifying an interface, the design challenge changes from “how does the user execute a task?” to “how does the user maintain context, control, and trust?” The UX patterns needed here are early and unproven, but they point toward more feedback loops, clearer diff presentations, and stronger undo or reset pathways.
Since a large share of Replit’s users are non-professionals building their first software, Hoang sees the opportunity to lower the floor still further while quietly lifting the ceiling. The tools are intended not to merely generate code, but to help the user behave like a better technologist over time.
The Future Creative Stack
Because AI blurs logical boundaries between once-distant phases of work, teams can no longer rely on a linear assembly line of design handoff and engineering build. Hoang observed that better tooling will need to be a single system that supports exploration without carrying forward fragmentation from different output formats or skill sets.
Looking forward, teams that treat AI as merely a cost-reduction lever tend to miss the bigger changes. The deeper shifts rest on whether a team reorganizes its workflow around the strengths of AI-supported creation, including better experimentation, greater personalization, and, ultimately, different kinds of artefacts than what existed long before.
Designing for the Unknown: Replit’s David Hoang on AI as a Creative Reset
David Hoang, Vice President of Marketing and Design at Replit, sees the current wave of AI as a familiar inflection point. He draws a direct line to the launch of the iPhone in 2007, when the rise of a new platform made everyone a beginner again. That reset, he argues, is the moment where new interface paradigms are born and where designers have the most to contribute.
Hoang discussed AI, the evolving relationship between design and code, and his approach to building with generative tools in a conversation with Figma Product Manager Mihika Kapoor. The discussion was published as part of Figma's The Prompt magazine.
From Dynamic Interfaces to a Single Discipline
For Hoang, the future of product design is not static. He envisions a shift toward what he calls "dynamic interfaces," where the presentation is not fully predetermined because experiences are becoming multimodal and multi-form. He connects this trend with other emerging technologies, arguing that AI and spatial computing are converging rather than developing along independent tracks. To meaningfully shape this future, designers will need to relinquish some of their instinct for control over the final interface.
This shift also erodes the traditional boundaries between design and engineering. Hoang believes AI’s primary role is augmentation, which will enable practitioners in either discipline to work across both. He predicts that engineering and design are on a path to becoming one tightly woven discipline. At Replit, this philosophy is centered on building what the company calls an Artificial Developer Intelligence (ADI) rather than pursuing Artificial General Intelligence. The strategy rests on two pillars: collaboration and AI, aiming to give users more autonomy and productivity.
From Learning to Shipping
Hoang positions Replit's ADI as a multi-faceted tool. It can manifest as simple code completion or evolve to augment organizational intelligence by understanding team context and improving collaboration. He frequently refers to the platform as a "technical co-founder" for users who want to move beyond merely learning to code. The goal is acceleration—getting people from the learning phase to shipping a product and building a business as quickly as possible.
He highlights the case of Priyaa Kalyanaraman, a participant in a non-technical hackathon track. Using Replit and AI, she built a platform that injected documents with personality using animated GIFs, playful language, and text-to-speech; AI wrote 100% of the code. The project won the hackathon, and Kalyanaraman went on to start the company Lica and secure funding.
Centralized Strategy, Experimental Interfaces
Replit’s internal structure for AI development is a hybrid model. A dedicated, centralized AI team handles the training of the company’s large language model and conducts product development research to better understand user needs with AI. While this team is central, Hoang notes that everyone across the company is thinking about how AI applies to their specific roles.
This structure creates a clear divide in how work is approached. For underlying infrastructure like APIs, services, and language models, Hoang argues that experimentation is not an option; these components need to be built thoughtfully and at scale. In contrast, there is significant room to experiment with the interface layer, since he believes the defining paradigm for AI interfaces has yet to be established. A large portion of the company’s work is focused on identifying that paradigm and anticipating user response.
Designers Are "Built for This"
Hoang is direct about the anxiety AI creates in the current job market but pushes back against the idea that it eliminates the designer's role. He asserts that designers will always have purpose because there will always be things that need to be designed. The crucial caveat is that the literal definition of "design" will evolve, and practitioners must be open to that change. He argues that when technology automates tasks, it paradoxically requires more critical thinking and creativity to envision what comes next.
He frames the challenge of designing with AI as nothing new: "Designing an AI is just systems design," he says. "You're doing the right things already, you just need to apply it to a different technological challenge."
Advice for Getting Started
For designers looking to enter the AI space, Hoang’s primary advice is to make time to play with products. This hands-on exploration is how one discerns quality AI use cases from features where AI is simply an afterthought. Understanding what AI can augment, automate, and replace is critical, he says, and it highlights the importance of contributing where AI cannot.
He believes the ultimate barrier for AI is not technological advancement but human adoption. The most innovative product will fail if individuals and companies are unwilling to learn the new behaviors it requires. He admits that the pace of change can induce imposter syndrome, but he reminds his team that the reset of the playing field is an opportunity to experiment, since no one knows yet what will be permanent.
This philosophy extends to his hiring and team management at Replit. The team favors multidisciplinary designers who are technically deep enough to code, prototype, review pull requests, conduct research, and run workshops. The expectation is that each designer has the output of three to five of their peers, and that product managers and engineers share this strong design taste and centricity.
Finally, Hoang connects this all back to community. In contrast to those who follow instructions step-by-step, he identifies as a person who wants to "open the thing and figure it out," build, ship, and iterate immediately. Successful communities, he notes, create a sense of cohort—a balance of accountability and fun that fosters a drive to keep learning.
