Smashing Conference Freiburg 2025

Two days in Freiburg, back-to-school vibes, and a fine lineup of talks that brought the web back to essentials: solid foundations, inclusive experiences, and interfaces that breathed better.

Brad put Atomic Design back into context: it’s neither a magic wand nor a dogma, but a shared language to align design and engineering around components, tokens, and open standards. The real work was organizational: reducing unnecessary variability (50 grays, 4 fonts), clarifying governance, and connecting the design system to product goals. He highlighted the momentum around design tokens (“subatomic”) and initiatives like Open UI to bring patterns closer to the web’s primitives.

The key message: a design system lives if it’s useful and interoperable. Favor modularity (tokens, primitives), actionable docs, and strong defaults. For multi-brand contexts, think “core + extensions,” not cloning. And stay pragmatic about stacks (Web Components vs React): optimize for durability, not hype.

Emily invited us to design for the “unhappy” paths: errors, emotional states, and real-world constraints. She proposed “lenses” to zoom in and out, identify risks, give control back to users, and connect user impact to business impact. Adding friction can be desirable when it serves clarity, safety, and ethics.

Practically, introduce risk early (research, prototyping), document what can go wrong, and craft messages that support rather than embellish. To persuade, measure the cost of ignored unhappy paths: support load, churn, compliance. Better a slightly slower, clearer flow than superficial “delight.”

Nick pinpointed the core flaw of dashboards: they often don’t answer basic questions (“is this good or bad?”, “what should I do?”). Without context, thresholds, prioritization, and action signals, the interface is decorative. Users say “too busy” when the value isn’t obvious.

The recipe: include targets, deltas, annotations, and highlight items that require action. Strip charts to the essentials, ban ambiguity (axes, units), and provide just-in-time explanations. The goal isn’t total freedom to assemble a wall of charts, but to support concrete decisions.

Ana proposed a realistic approach to legacy CSS: map it (tools like cssstats), set measurable goals (weight, duplication, scope), and modernize in layers rather than rewriting everything. Cascade layers can “park” old styles while introducing healthier patterns.

Prioritize low-risk quick wins, isolate side effects with visual regression tests, and progressively introduce conventions (naming, scopes, variables). If migration is needed, think “strangler pattern”: new clean zones, old encapsulated ones, with metrics guiding the pace.

Kardo shared his journey as a visually impaired designer: accessibility isn’t an add-on, it’s a stance. His stories reminded us that an inclusive interface must withstand real contexts: strong contrast, visible focus, non-visual alternatives, robust navigation.

He called for normalizing asking for help, involving affected people in teams, and using AI judiciously: useful for personalization and testing, risky if it “smooths” interfaces at the expense of clarity and control.

Matthias urged us to “paint with the web”: abandon the myth of static pixel perfection and embrace a fluid medium. Concretely: fluid typography and grids, resilient components, progressive enhancement, and designing in the browser.

He advocated for more living deliverables (prototypes, “design tokens + stories”) and hybrid teams where dev and design meet earlier. The goal: adaptive, accessible, durable interfaces without over-specifying what the web already does well.

Scott laid out practices that move the metrics: optimize LCP (no loading=lazy on the key image, fetchpriority=high), inline critical CSS, self-host fonts, use HTTP 103 Early Hints, and monitor via CrUX/SpeedCurve/WebPageTest. Performance is discipline, not isolated hacks.

He also reminded us to tackle third parties and delivery (caching, navigation speculation) before micro-optimizing. And yes, many gains come from clean server-side HTML; frameworks should cooperate rather than hide complexity.

Oliver brought philosophy into practice: the “knowledge of making” (Vico) and respect for attention. Start with readability (type, hierarchy), say something useful, and cut the superfluous. A good interface is first a good text, well presented.

He also questioned the lure of “cool” and noise: design improves when anchored in clarity, relevance, and ethics—not one-upmanship.

Manuel showed how “new CSS” is liberating: functions, perceptual color, custom properties, modern patterns—if we’re willing to challenge inherited habits. On a fresh project, he reevaluated every line and proved that well-thought simplicity replaces layers of tooling.

Applied to legacy, his approach encouraged documenting CSS decisions, exposing variables, and building components truly independent of the tool (yes, even if you like Tailwind, keep understanding CSS).

Kevin put things back in perspective: to serve everyone, you have to invest in research, even with limited means. Better identifying critical segments saves money later, avoids “enshittification,” and guides trade-offs in the impossible cheap/fast/good trio.

He offered realistic strategies for small teams: tight scoping, frequent testing, impact-risk prioritization, and a clear stance on product ethics. AI helps, but it doesn’t replace listening or method.

Cassie live-coded a GSAP game, showing that animation and accessibility can coexist when you honor user preferences (reduce-motion), provide alternatives, and manage focus well. The show was a reminder that interactivity isn’t an end in itself: it serves meaning when it stays sober and controllable.

It was also a plea for learning by doing: prototype, test, iterate—and keep the experience readable when the visual effect quiets down.

This 2025 edition told a single story, from design systems to performance to accessibility: get back to reality. Reality is constraints (legacy, budgets, org), humans (emotions, contexts, disabilities), and browsers that evolve faster than our processes. The middle path is to be demanding about fundamentals (performance, readability, semantics, accessibility) and pragmatic about the rest (tools, frameworks, trends). In short: fewer illusions of control, more clarity, measurement, and intent. We’re heading back to Lausanne with a few simple to-dos: audit our LCP, clean up historical CSS, write messages that truly help, and design dashboards that answer “So, what do I do now?” The web is a flexible medium—let’s be just as flexible.