Tuesday, November 17, 2026vbrunetti.com · Personal Edition

The
Morning
Brief

Tuesday, November 17, 2026

AI tools are getting better at satisfying functional briefs but worse at differentiation — a tension that sits squarely at the intersection of design systems and creative judgment. Meanwhile, the 'human in the loop' assumption is being empirically stress-tested, with data suggesting that the value designers bring must be more intentional and specific than mere oversight. The implication for design leadership: the craft and systems work that produces distinctive, contextually appropriate output is becoming the irreplaceable layer.

AI Design Tools Have Solved Competence. They Haven't Solved Character.
Design Tools

AI Design Tools Have Solved Competence. They Haven't Solved Character.

Jakob Nielsen

A benchmark across 15 frontier AI design models finds they satisfy 90–99% of functional brief requirements — a remarkable threshold — yet their outputs are visually indistinguishable from one another and lack the domain-specific variation human designers produce. This is the clearest empirical framing yet of where AI ends and design systems leadership begins: AI can execute a brief, but it cannot encode a point of view. The implication for design systems work is significant — your tokens, component decisions, and brand logic are precisely what transforms competent AI output into something that actually belongs to your product.

The 'Human in the Loop' Assumption Doesn't Survive a Meta-Analysis
AI

The 'Human in the Loop' Assumption Doesn't Survive a Meta-Analysis

Jakob Nielsen

Across 420 studies, AI beats humans alone 68% of the time — and human-AI teams perform worse on average than the stronger solo performer. The exception holds only where humans possess information the AI genuinely lacks, which reframes the designer's role from reviewer to context-provider.

Diamond Prompting: Parallel Design Is Now a Solo Workflow
Design Tools

Diamond Prompting: Parallel Design Is Now a Solo Workflow

Jakob Nielsen

Generating ten deliberately divergent AI designs before converging on the best — 'diamond prompting' — costs roughly $53 with premium models, making parallel design exploration accessible to individual practitioners. The technique also exposes a methodological flaw in how AI design tools are benchmarked: testing only a model's first idea hides most of its range.

What 15 Hours of Shadowing Developers With AI Agents Revealed
Ways of Working

What 15 Hours of Shadowing Developers With AI Agents Revealed

Jakob Nielsen

Field research watching developers work with AI coding agents shows patterns no lab study captures: multi-tasking across 1–17 concurrent tasks, dispatching agents and returning later, and relying on team norms — not tool guardrails — to decide what needs human review. For anyone designing AI-assisted workflows, the gap between what users can do and what they actually do is the design problem.

AI design tools have cleared the competence bar; the new design moat is the system of decisions — tokens, patterns, brand logic — that makes output recognizably yours.

Jakob Nielsen →

If scientists with 7 hours of weekly AI savings are still underusing it for ideation and hypothesis generation, the highest-leverage creative applications remain largely untapped across knowledge work.

Jakob Nielsen →

The curse of knowledge is a design systems risk too — the longer a team lives inside their component library, the less they can see it as a new adopter or contributor would.

Jakob Nielsen

The 'seamful design' principle — selectively exposing system limits rather than hiding every seam — maps directly onto how AI-assisted design tools should communicate their constraints to practitioners.

Jakob Nielsen