Overview
Design engineer Maggie Appleton argues that product design and software engineering are closely related forms of problem-solving conducted through different materials. Her career—from cultural anthropology and illustration to frontend engineering, AI-product design, and GitHub Next—shows why strong designers must understand not only visual composition but also data structures, system behavior, performance, users, and cultural context. AI agents have transformed her workflow by making polished, live prototypes cheap enough to build, discard, and rebuild rapidly. They can translate Figma concepts into working interfaces and generate custom controls for tuning animation, color, spacing, and other variables directly in the browser. Yet they remain unreliable, spatially weak, aesthetically repetitive, and highly dependent on human direction. Consequently, paper sketches, familiar interface patterns, user research, and expert taste remain essential. Appleton believes the larger unresolved problem is organizational rather than individual: agents accelerate isolated developers, but software is still built by teams that need shared planning, decision ownership, context, and audit trails before implementation. Her broader conclusion is that AI should widen participation in software creation while preserving human craft, cultural awareness, data ownership, safety, and the ability to question what should be built in the first place.
Sections
Key Concepts
Terms used to explain the emerging relationship among design, engineering, AI agents, and personal software.
- Design engineer: a designer who understands the product’s technical architecture, works closely with engineers, and often implements interfaces while retaining responsibility for product language, behavior, motion, and visual quality.
- Product nouns and verbs: the concepts users can identify in a product and the actions they can perform on them; together they form the operational vocabulary through which users understand a system.
- Jig: a temporary, task-specific control interface—such as sliders, color pickers, or motion controls—generated to tune variables directly inside a live prototype.
- Capability gaslighting: the disorienting experience in which a model appears highly capable after an impressive result, then fails unpredictably on another task or under different context.
- Digital garden: a public body of writing that can contain explicitly labeled unfinished ideas and be revised over time through stages such as seedling, blooming, and evergreen.
- Home-cooked software: small software made for oneself, one’s family, or immediate community, analogous to preparing a homemade meal rather than purchasing a mass-market product.
- Barefoot developer: an advanced local user who employs accessible tools or agents to build software for community needs without being a highly paid professional developer.
Higher-Level Implications
Patterns that emerge across Appleton’s experiences with anthropology, design, implementation, and AI agents.
- As implementation becomes cheaper, problem framing, shared decision quality, and validation become a larger proportion of valuable software work. AI shifts effort upstream rather than eliminating the need for expertise.
- The most useful AI design workflow is hybrid: nonverbal exploration on paper, approximate composition in Figma when helpful, live browser experimentation through generated controls, and human evaluation throughout.
- Interfaces mediate between two very different domains: model systems optimized for text, weights, tools, and computation, and humans who reason through space, culture, memory, materiality, and social context. Better agent products must support translation between these domains.
- Aesthetic quality is culturally situated and changes over time. Once agent-generated visual conventions become recognizable, differentiation will depend on human taste and the creation of new visual languages rather than flawless imitation of current trends.
- Anthropological immersion offers engineers a path to greater leverage: becoming closer to users and their real environments can reveal needs and constraints that code generation alone cannot discover.
Important Contrasts
Explicit contrasts used to clarify roles, tools, and emerging workflows.
- Design and engineering use different immediate materials but share the same core activity: defining problems, evaluating alternatives, building models, and validating solutions.
- Paper offers speed, persistence, and pre-verbal spatial exploration, whereas agent prompts require ideas to be translated into language before implementation.
- Figma helps establish approximate visual form, color, hierarchy, and scale, while a live browser prototype exposes responsiveness, rendering, scrolling, motion, and experiential smoothness.
- Traditional handoffs separate polished design artifacts from implementation constraints; design engineering keeps the designer close to code, data, and runtime behavior.
- GitHub’s production designers protect a mature system used by millions, while GitHub Next explores riskier prototypes with lower production constraints and a longer time horizon.
Practical Lessons
Actionable conclusions grounded in the guest’s work and experiments.
- Begin ambiguous visual or architectural thinking with the lowest-friction medium available; boxes and arrows can reveal structure before language is ready.
- Move interface concepts into a real browser early enough to assess responsiveness, scrolling, rendering, motion, and data behavior.
- Ask agents to generate temporary controls for uncertain variables so designers can tune the live system directly instead of repeatedly editing static specifications.
- Preserve familiar interface primitives unless research shows that a new interaction model produces enough benefit to justify its learning cost.
- Treat agent decisions as shared artifacts with an owner, relevant evidence, context, and a durable record that teammates can inspect later.
- Use agents as patient tutors across disciplinary boundaries: designers can learn systems and state, while engineers can learn typography, usability testing, and user research.
- Verify model output even after impressive successes because capability varies with prompts, context, randomness, and model changes.
- Support nonprofessional developers with local-first, secure, durable foundations rather than expecting prompt-generated code alone to produce safe software.
Memorable Quotes
Statements that capture the interview’s central arguments in the original Arabic transcript.
- «تصميم المنتجات، عندما نتحدث عنه في مجال التقنية، هو في الواقع تصميم برمجيات، وتصميم البرمجيات هو هندسة. إنها في الواقع نفس المهارة.»
- «عليك أن تفهم المواد التي تستخدمها في أي دور تصميمي.»
- «إنها مجرد طريقة لتفريغ الأفكار قبل تحويلها إلى لغة.»
- «وبعد شهرين من التساؤل: "ما هي واجهة المستخدم الجديدة للذكاء الاصطناعي؟"، وجدنا أنها مجرد جدول.»
- «عندما نعمل بمفردنا مع الوكيل، يمكننا العمل بسرعة كبيرة. لكن العمل الجماعي أساس النجاح.»