Fast-Fashion Interfaces: Assigning Taste When Every Surface Is Temporary

In Fast-Fashion Software I argued that most tactical software should be disposable—the campaign dashboard, the seasonal microsite, the internal tool that dies when the question dies. That frame is still right for artifacts. The next layer is harder: the interface itself is becoming fast-fashion. Not the whole product. The surface between human and outcome—spawned for a user, for a request, for a Tuesday, then replaced when someone on the team clones it and inverts it into something better for their workflow.

Enterprises are not ready for this. They are still organized around permanent applications, shared design systems as law, and the fantasy that one interface serves every role. The future outside the outlets looks different: code-editor logic applied to business work—imagination in, working surface out—with assigned arbitrators of taste who decide what compounding quality means while everyone else ships tangential, temporary UI at the speed of thought.

The code editor was the preview

Watch how builders actually work. They do not open a fixed application and click through someone else's information architecture. They open an editor—or increasingly, a conversational workspace—and move from intent to artifact in tight loops: sketch, run, break, rename, ship. The interface morphs around the task. Files appear. Panels rearrange. A terminal slot opens because today's problem needs logs, not because the product roadmap said Q3.

That mental model is migrating to business operators who never called themselves engineers. Marketing ops spins a review surface for one client. Finance wraps a reconciliation flow that exists for eight days. Strategy drops an interactive brief because the steering committee asked a question the deck could not answer. The interface is not the company's official portal. It is the shortest path from this person's request to this outcome—and tomorrow the request changes.

I wrote in As Fast As You Can Think that the bottleneck moved from syntax to clarity of intent. The same shift applies to UX: when generation is cheap, the scarce skill is knowing what surface should exist for this moment, not polishing a global nav for the fifth year running.

Two speeds: cathedral and tangent

Not every interface should be disposable. Confusing the two modes is how enterprises will waste the next cycle. I see two distinct speeds:

• Cathedral interfaces. Magical, immersive, deliberate—planned the way a flagship store is planned. Brand worlds, client portals that carry trust, onboarding that must feel inevitable, public surfaces that sign your name. These deserve craft, rehearsal, accessibility, performance budgets, and owners who treat them like machine-readable brand infrastructure. They are rare. They compound.
• Tangent interfaces. Day-to-day workflow shells—internal, contextual, outcome-first. The UI is not the product; it is packaging for a decision. It may live for a sprint. It may be wrong by Friday and nobody should mourn it. This is the bulk of enterprise work, and pretending it deserves a six-month design review is how you lose to operators who shipped three versions before your kickoff meeting.

Hot work and deep work from Hot Work and Deep Work rhyme with this split: cathedrals are built in deep work; tangents are often hot-work surfaces—fast, functional, gone. The org that only builds cathedrals moves too slowly. The org that only builds tangents erodes trust. You need both, labeled honestly.

Assign arbitrators of taste

When anyone can spawn an interface, "everyone is a designer" becomes "nobody is accountable for quality." The fix is not central approval for every screen. The fix is named arbitrators of taste—people whose job is to hold the compounding bar for the layers that matter:

• What a cathedral must feel like — voice, motion, accessibility, trust cues, the non-negotiables when the interface carries the brand.
• What a tangent must satisfy — legibility, correct data, audit trail, sunset date—not pixel perfection.
• What gets promoted — when a throwaway workflow proves it has a second season, who elevates it to long-lived infrastructure and who pays the maintenance tax.
• What gets killed — the equally important power. Fast fashion requires disposal. Arbitrators who only bless and never bury create the cruft layer that killed every prior generation of internal tools.

This is not a committee. It is a small, explicit role—often overlapping with brand, design ops, or a senior operator with taste and veto power. The mindset case in The Battle of Mindsets applies: when output is abundant, hire and promote for judgment, not for who owns Figma.

The interface must change with the user and the request

Personalization used to mean "show their name in the header." That era is over. The serious version is structural: different operators see different compositions for the same underlying capability because their requests, permissions, and cognitive load differ.

A regional lead needs exception queues and narrative context. A analyst needs raw tables and export. An executive needs three numbers and a confidence band. One backend, three tangent interfaces—possibly generated in session, not shipped as three permanent nav items. The harness from The Harness Is Becoming More Important Than the Horse is what persists; the chrome is fashion.

Enterprises still buy monolithic suites because procurement loves one throat to choke. Operators inside those suites already rebuild tangents in spreadsheets, Slack workflows, and sidecar tools. AI accelerates that behaviour from workaround to default production mode. The interface becomes a function of (user, request, moment)—not a static map in a wiki.

Teams will clone, invert, and remix each other

Here is the social dynamic nobody puts on the roadmap. Person A ships a beautiful, functional tangent for their squad. Person B clones it, inverts the layout, swaps the mental model, ships a variant tuned to a adjacent process. Person C forks it into something ugly that happens to solve the CFO's question faster. This is not sabotage. This is the workflow.

Game culture understood this decades ago. Tower-defense maps, custom scenarios, total conversions—communities where distinct authority and style matter (who built the canonical version) but remix is expected. You do not sue someone for making another lane layout; you argue about taste, balance, and whether the fork respects the spirit of the original. Enterprises have no equivalent etiquette yet. They have IP panic and design review boards built for a world where copying UI was expensive.

Copying UI is now cheap. The organizational response cannot be lockdown. It must be: credit, lineage, arbitrators, and sunset. Fork the workflow, name the parent, set a tombstone date, let the taste arbiter promote the winner. The alternative is shadow IT with better typography.

The vintage loop

Fast fashion does not only move forward. Culturally it loops—thrift, archive drops, the nineties returned because the cycle needed contrast. Software will behave similarly. Teams will rediscover "vintage" stacks: the minimalist internal tool from 2024 that worked, the runbook UI everyone hated until the new suite got bloated, the terminal pattern that came back because tangents got too clever.

That rhythm is healthy if you expect it. The compounding layer—skills, knowledge, brand rules, data contracts—stays in the archive. The interfaces rotate like seasons. Enterprises that treat every retired screen as failure will miss the point. Disposal makes room for the return of what actually worked.

Not everyone will be comfortable

I would say this plainly. A large fraction of the enterprise is built for stable applications, predictable change windows, and the psychological safety of "the system we trained people on." Tangential, ever-changing interfaces feel like chaos to them—and sometimes they are right, when tangents touch compliance, money, or customer trust without a harness.

The leadership move is not to force everyone into editor brain overnight. It is to partition comfort zones: cathedrals and governed domains for those who need stability; tangent factories with taste arbitrators and clear rules for those who ship at pace. Mixed messages—demanding innovation while punishing disposal—produce the corporate impotence cycle I wrote about in The Corporate Impotence Cycle: more slides, less proximity to the work.

What to do this quarter

• Name two taste arbitrators (brand-facing and ops-facing if you must split). Give them promote/kill authority on interfaces, not just critique rights.
• Label every new surface cathedral or tangent at creation. Default tangent for internal workflow. Default cathedral only when the brand signs its name.
• Require tombstone dates on tangents. Promotion to long-lived is a deliberate act with an owner—same discipline as Fast-Fashion Software.
• Allow fork lineage — clone, invert, credit. Measure outcomes, not sameness.
• Invest in the harness — APIs, permissions, audit, shared data—not in one eternal UI shell.

The future outside the outlets is not one perfect application. It is millions of temporary interfaces, a small set of cathedrals, and humans assigned to taste—not because they are the only ones who can build, but because someone must decide what compounds and what goes out of season. Prepare for the swift change. The code editor was never just for code. It was the template for how imagination becomes interface—and how interface becomes disposable the moment the request moves on.

Related: Fast-Fashion Software, Taste, Memory, and the Loop That Compounds, In-House Excellence vs Agency Fuel, and The Battle of Mindsets.