UX Collective: Designing for AI, the permalink problem, vibe designing

In a June 2026 edition of his UX Collective newsletter, Fabricio Teixeira—founding editor of one of the largest UX publications online—examines three ideas circulating in design communities that he argues deserve more systematic attention than they have received.

The state of designing for AI

Teixeira observes that AI tools have accumulated extraordinary capability over the past few years while design conventions for how to present that capability to users have largely not followed. The tooling continues to be built while it is already in use, which means teams are making interface decisions without shared patterns to reference. He describes this as conditions where the “ground is soft, shifting like tech quicksand”—a useful framing for teams that are building AI features and finding that existing design patterns from web or mobile don’t transfer cleanly.

The permalink problem

Most AI conversational interfaces produce no shareable output. The conversation exists only in the browser session where it happened. Unlike a web article, a document, or even a spreadsheet, there is no URL to send a colleague, no record to audit, and no way to link back to a specific exchange. Teixeira frames this not as a minor UX detail but as a structural issue for workflows built around AI-assisted work. Version history, review processes, and institutional memory all depend on reference-ability that current chat interfaces don’t provide. The question of where AI outputs live and how they can be referenced is one that design teams building AI-forward products will need to answer.

Vibe designing

Teixeira positions “vibe designing” as the design profession’s equivalent of vibe coding—the practice of using natural language prompts to generate functional software through iteration, without explicit specification. Applied to interface work, vibe designing starts from a feeling or aesthetic direction and iterates through AI-generated outputs rather than working from detailed briefs or component-level decisions. He notes that this produces results that can be fast and directionally useful while also being difficult to hand off to engineers or validate against systems thinking.

Who it is useful for

The article synthesizes these threads rather than resolving them. It is most useful for product designers, design leads, and UX practitioners who want a conceptual frame for the interface problems AI products are generating—particularly those wrestling with how AI outputs can be tracked, shared, or built upon across a team. The value is in naming patterns that are already present in how teams work with AI tools but that haven’t yet been formally described.