🎨 Design & UX

AI in design, UX role shifts, design tooling.

RepoVoltAgent

awesome-claude-design: 68 DESIGN.md files to stop AI apps looking identical

A curated set of 68 `DESIGN.md` files, each modelling the visual design system of a well-known brand (Linear, Vercel, Notion, Stripe, and more), meant to be dropped into Claude Design or a coding agent to scaffold an on-brand starter UI instead of a generic default.

Each file follows nine standardised sections — colour roles, typography, components, layout, depth, constraints, responsive behaviour, and agent prompts.
Uploading one to Claude Design produces CSS variables, font substitutes, and a working starter page in one pass.
Explicitly framed as inspirations modelled on public patterns, not official brand kits — usage still needs to respect each brand's trademark rules.

added by Radar • 12th Aug 2026

BlogMatt Rothenberg

You Don't Have a Design System

An experiment-backed essay arguing that a component library and token set aren't a design system if they don't also carry forward shared decisions about how the product should behave.

Three AI agents given the identical library (shadcn/ui on Base UI), token layer, and app shell each built a different settings page — some behind tabs, some as cards, some as long forms — despite passing the same token linter.
The failure wasn't in the components; it was in composition — nobody had made a canonical decision about what a settings page *is*, so each agent answered the question separately.
The real missing work of a design system is choosing an answer, making it canonical, and putting it where the next builder (human or agent) will find it.

added by Adam Tomat • 5th Aug 2026

BlogUX Collective • Patrick Neeman

Information Architecture Is the Foundation AI Is Starving For

Argues that information architecture, long underfunded because its failures were diffuse and hard to measure, now has a concrete business case because AI systems expose structural flaws as measurable token costs and hallucinations.

Metadata, taxonomies, and content typing let AI systems tell authoritative information from anecdotal — without them, "a price, a policy, a deprecated note, and a customer quote can read as similar strings of text and mean opposite things."
Humans forgive vague labels by using judgment; AI pattern-matches over messy structures and confidently reproduces errors at scale, turning isolated mistakes into systematic ones.
IA can now be funded through measurable AI outcomes — retrieval accuracy, hallucination reduction, agent reliability — rather than abstract UX arguments.

added by Adam Tomat • 28th Jul 2026

BlogHarvard Business Review • Melchior Tamisier-Fayard, Theodoros Evgeniou & Anne-Laure Fayard

Design AI Systems That Actually Strengthen Human Reasoning

An HBR piece arguing that AI systems must be deliberately designed to preserve employees' critical-thinking capacity, not just their output speed.

Draws on research from management, cognitive science, and human-computer interaction showing people tend to accept AI outputs without enough scrutiny.
Frames the erosion of critical reasoning as an organizational risk, not just an individual habit — it undercuts the innovation and agility companies need AI to enable.
Argues for intentionally architecting AI tools to prompt questioning and challenge-seeking, rather than defaulting to passive acceptance of machine-generated recommendations.

added by Radar • 27th Jul 2026

BlogNielsen Norman Group • Tony Alicea

UX-Context Design: Using UX Knowledge to Steer AI-Generated Design

Nielsen Norman Group argues that as PMs and engineers increasingly generate designs with AI, UX's core deliverable needs to shift from wireframes toward structured, machine-readable context.

Research output should move from human-readable artifacts (personas, journey maps) to context AI systems can consume directly.
Points to Google's open-sourced DESIGN.md format as a model: exact machine-readable values (colors, spacing) paired with human-readable rationale, so AI knows what a value is *for*, not just its number.
Frames this as continuous curation, not a one-off handoff — the context file lives alongside code and gets updated as research evolves and AI output is monitored.

added by Radar • 25th Jul 2026

BlogUX Magazine • Tushar Deshmukh

Your First 30 Days With AI in Your Design Workflow

A four-week plan for designers to deliberately test AI alongside their existing process, week by week, rather than adopting or dismissing it wholesale.

Week 1 sets a baseline by running AI alongside current practice for comparison, not as a replacement.
Weeks 2–3 deepen the promising use cases and then apply AI carefully to more sensitive research work.
Week 4 turns the accumulated evidence into a personal map of where AI helps, where to use it selectively, and where to avoid it — human judgment stays essential for context and nuance AI still misses.

added by Radar • 25th Jul 2026

BlogBusiness Insider

AI Raises the Bar for What Design Candidates Must Show

An interview with Figma VP of Product Design Noah Levin arguing that AI has raised, not lowered, the bar for design job candidates.

Candidates are now expected to show interactive, clickable prototypes rather than static screenshots, to prove real design thinking.
Levin wants to see AI used as a collaborator, not a shortcut — candidates should show discarded ideas alongside the final work, not just polished "AI slop."
Practical takeaway for hiring managers: probe whether candidates understand platform conventions and how they actually used AI during the process.

added by Radar • 20th Jul 2026

RepoTogether AI • Nutlope

Hallmark: An Open Design Skill That Fights AI-Slop UI

An open (MIT) design skill for Claude Code, Cursor, and Codex that generates interfaces deliberately built to avoid the generic "AI-slop" look.

Picks a page macrostructure and one of twenty themes, then runs 57 anti-pattern "slop-test" gates before output, plus a self-critique pass.
Covers four modes: build a new UI, audit existing code, redesign a layout, or study a design reference.
Installs with `npx skills add nutlope/hallmark` or by copying files directly into a Claude Code, Cursor, or Codex project.

added by Radar • 16th Jul 2026

BlogUX Collective • Aurélie Radom

Taste Cannot Be Delegated

An argument that as AI makes production and preference-prediction cheap, design judgment — knowing *why* something is the right choice, not just what people tend to pick — becomes the scarcer and more valuable skill.

Draws a line between preference (what people choose, which AI can predict from patterns) and judgment (why it's the right choice, which requires context and conviction).
Design-by-committee dilutes decisions toward the average as more stakeholders weigh in, and accountability quietly disappears as collaboration replaces ownership.
Metrics and optimisation refine what already exists; genuinely new formats come from someone exercising judgment, not from data validating a safe choice.

added by Adam Tomat • 14th Jul 2026

BlogUX Collective • Pablo Stanley

The Terminal Became My Canvas

A designer's account of shifting their primary creative tool from Figma/Illustrator to the terminal, working directly with AI coding agents to build products instead of mockups.

Reports building in a day what used to take weeks, crediting the shift from visual tools to text-based, agent-driven work.
Working in the terminal lets them explore many ideas in parallel instead of committing to one direction early, changing how fast concepts get validated.
Frames it as a genuine productivity paradox: real gains in speed and output, alongside real drawbacks from leaving familiar visual design tools behind.

added by Adam Tomat • 14th Jul 2026

BlogUX Collective • Zeeshan Khalid

When Design Becomes the Organisation, and AI Becomes the Design

An essay arguing that AI isn't just a design tool anymore — as it reshapes how organisations operate, design itself becomes the mechanism by which the organisation is designed.

Poses the core paradox: what does it mean when the entity doing the designing is itself increasingly designed by machines.
Pushes past workflow efficiency into questions of identity and agency — what "being a designer" means once AI is integral to all aspects of the practice.
Frames the shift as organisational, not just tool-level — AI restructuring how design happens changes the mechanism through which the wider organisation is shaped.

added by Adam Tomat • 14th Jul 2026

BlogMC Dean

A Constellation of Emerging Design Roles

An argument that traditional design job titles aren't disappearing so much as fragmenting into AI-era specializations, using "adaptive radiation" as the framing for how designers can branch into new roles.

Hybrid titles — Design Engineer, AI Experience Designer, Computational Designer — are showing up on job boards, signalling that orgs now expect AI fluency as a baseline design skill.
Reinvention isn't required: existing strengths (systems thinking, craft, research, taste) each point toward a specific emerging niche that needs just one added capability.
Real working knowledge of how AI behaves matters more across every niche than prompting tricks or templates.

added by Adam Tomat • 12th Jul 2026

ToolNameThat

NameThat: The Visual Dictionary of UI

A visual dictionary that lets you describe a UI element in plain language (or point at it) and get back its real technical name, so you can prompt coding agents precisely instead of vaguely.

Covers native-app and web-in-a-shell vocabulary separately, since the same-looking element often has a different name on each platform.
Includes comparative guides (e.g. AppKit vs SwiftUI naming) and a translation table across 60+ UI elements.
Double-press any word on the site for a plain-English definition; command-palette search (⌘K) for quick lookup.

added by Adam Tomat • 12th Jul 2026

BlogLuis Ouriach

Design System Architecture: Theming, Experiments, and Keeping It Maintainable

A practical look at the architectural decisions mature design systems face when scaling across teams and platforms: managing experiments, theming, web/app library splits, and token granularity.

Treats component sprawl as a governance problem, not a technical one — Figma's Extended Collections let value-level experiments run independently, but structural changes need a formal decision process to avoid fragmentation.
Recommends keeping seasonal/brand variants as separate, toggleable libraries distinct from modes (light/dark), and warns that splitting web and app libraries is harder than it looks — don't do it until there's both structural necessity and dedicated ownership.
On tokens, start semantic and add component-level specificity only when experiments demand it; on AI tooling specifically, argues today's single-file structures suit AI generation but multi-file structures still scale better for human teams, so prioritise team workflow over premature AI optimisation.

added by Adam Tomat • 11th Jul 2026

BlogTechCrunch • Ivan Mehta

Figma Acquires the Bud (ex-Orchids) Team to Build AI Agents Into Design

Figma is acquiring the team behind Bud, a YC-backed startup that pivoted from vibe-coding app builder "Orchids" into an AI agent that browses the web and writes code.

Both Bud and the original Orchids product shut down 18 July 2026; existing users must migrate projects before the deadline.
The move is part of Figma pulling an AI coding-and-prototyping layer onto its canvas, following recent web-app-creation and coding-assistant integrations.
Orchids previously had security issues that left user-built apps open to attack, per earlier BBC reporting.

added by Radar • 8th Jul 2026

Repodanielalbinsson

An AI Agent That Reviews Pages for A11y and Design-System Drift

An open-source multi-agent tool that audits a live, rendered URL for accessibility, UX heuristics, and design-token drift, then produces a severity-ranked report — including as GitHub PR comments.

An orchestrator runs three specialist subagents in parallel: an accessibility auditor (axe-core), a visual-heuristics critic (vision models), and a design-system checker validating against design tokens.
Audits the actual rendered page in a real, headless-Chrome browser rather than reading source code.
Can be pointed at a pull request and post its findings directly as PR comments.

added by Radar • 4th Jul 2026

BlogChristine Vallaure

You Design It, Then What: A Clear Map of the Figma-to-Code AI Mess

A breakdown of why the Figma-to-code AI pipeline confuses people: it's really four layers of different maturity (MCPs, markdown memory/rules, skills, and design-to-code mapping), and demos only ever show one clean layer under perfect conditions.

Match the stack to your team size — solo designers need only pipes and notes, small teams need shared notes and habits, and all four layers are only worth it for large design-system teams.
"Stop measuring yourself against a setup built for a company of hundreds."
Before any of it works well, Figma files need real structure — styles, variables, auto-layout, actual components — since good input is the cheapest upgrade available.

added by Radar • 4th Jul 2026

BlogChristine Vallaure

What Are Hypertokens? The Layer Between Tokens and Components, Rebuilt for Agents

Explores hypertokens, a proposed design-token layer (an exploration by Jake Albaugh's team at Figma) that bundles related style properties into one named source of truth compiled into every tool's format.

Solves token drift: bundles like typography (family, size, weight, line-height) currently live separately in CSS, Figma, and iOS, and go out of sync.
Gives agents a single semantic handle (e.g. `Surface.brand`) instead of fifteen loose values to guess how to connect.
Not yet shipped — still an exploration, but worth adopting semantic design-system practices now to align with it later.

added by Adam Tomat • 3rd Jul 2026

BlogDesign Week

Koto Rebrands Stack Overflow for the AI Era

Coverage of Koto's rebrand of Stack Overflow, repositioning it from a static Q&A archive to a "living truth engine" built on verified human knowledge, in response to AI.

The brand case rests on Stack Overflow's 83M+ community-contributed Q&As as human-validated data AI models can't replace.
Design routes were shared publicly and put to a community vote at WeAreDevelopers World Congress before launch.
Koto built a custom generative tool, made with Claude, so teams can vary composition, motion, and colour while staying on-brand.

added by Adam Tomat • 2nd Jul 2026

BlogCreative Boom • Tom May

The State of the Creative Industry 2026

A survey of 882 creatives worldwide on money, burnout, and AI, finding widespread AI adoption but little enthusiasm for it.

69% report burnout in the past year (77% among mid-career creatives), while nearly half of self-employed creatives earn under £30,000.
86% use AI tools professionally, but only 10% think AI is having a positive effect on the industry — adoption driven by necessity, not belief.
What people say would actually improve their working lives is networking and community (57.5%) and mentorship (53%) — new tools trail at 31%.

added by Adam Tomat • 2nd Jul 2026

BlogPenpot • Juan de la Cruz García

Why Fast AI UI Generation Still Falls Short for Enterprise Design

A design-tool maker's argument that fast AI UI generation (like Claude Design) misses what enterprise teams actually need: governance, integration, and data ownership, not just speed.

Proprietary, closed environments raise compliance and vendor-lock-in concerns for regulated industries.
Rapid prototyping doesn't translate into maintainable component libraries, design tokens, or governance at scale.
Proposes MCPs as the fix: giving AI agents access to shared design components/tokens so output aligns with an existing design system automatically, instead of needing to be rebuilt by hand.

added by Radar • 2nd Jul 2026

BlogUX Collective • Kike Peña

You're Missing the Point: This Is Your Real Value in Tech Companies

A Chief Design Officer's argument that professional value in tech now comes from applying judgment and strategy on top of AI tools, not from choosing between automation and human skill.

Real value is the ability to apply criteria, knowledge, and vision to a problem — the tools are secondary to that judgment.
Companies that lean on AI output without human judgment end up with generic, indistinguishable products.
Using the technology well requires understanding its benefits and limits first, rather than chasing hype trend by trend.

added by Adam Tomat • 2nd Jul 2026

BlogUX Collective • Dolphia

Figma Just Made Your Design System Debt Everyone's Problem

Resolved via search — verify before relying on it. Figma's new "Check designs" tool scores a file against your design system and surfaces every drift — hard-coded values, failed contrast, detached components — turning design-system debt from an abstract complaint into a visible number.

A visible drift count gives design system work leverage in reviews it used to lose quietly.
Within days of launch, designers found it flags intentional exceptions (annotation colors, dev notes, disabled states) as violations, with no way to mark them as deliberate.
The count is leverage, not a strategy — someone still has to separate real signal from false positives.

added by Adam Tomat • 1st Jul 2026

BlogUX Collective • The UX Engineer

Why Systems Thinking Is Becoming the Most Important UX Skill

As AI agents increasingly call app functions directly without a screen in between, the argument that a designer's job is shifting from shaping screens to shaping systems.

AI tools are commoditising polished visuals and design-system production, so judgment and taste matter more than execution speed.
Platform-level agent integrations (e.g. Siri invoking app functions without launching the app) make cross-platform systems thinking a core skill, not an afterthought.
Recommends trading screen-based wireflows for Job Stories, state charts, and flow diagrams to describe how apps behave across contexts.

added by Adam Tomat • 1st Jul 2026

BlogFigma • Georgia Rust & Rodrigo Davies

Figma's Design Agent, Now With Custom Tools and Greater Context

Figma's design agent gains custom tools, richer context sources, and reusable skills.

Build custom reusable generative plugins on the canvas.
Attach files and reference Figma files / web / MCP connectors for context.
Turn frequent prompts into shareable slash-command "skills".

added by Adam Tomat • 29th Jun 2026

BlogVercel • John Phamous

Teaching agents product design at Vercel

Vercel's repo-based skill that encodes product and design decisions for coding agents.

Design decisions encoded as a versioned skill.
Keeps agent-built UI consistent with the product.
The skill lives in the repo, next to the code.

added by Adam Tomat • 26th Jun 2026

VideoWes Bos

Browsers and UIs are dead. Everything is chat (Wes Bos)

Wes Bos's JS Nation Amsterdam 2026 talk on agentic interfaces and Web MCP — pushing back on "chat replaces all UI".

Frames the "clicks vs clankers" tension: build AI into your site, or build your site so agents can use it directly.
Covers Web MCP as the mechanism letting agents use a site's existing capabilities instead of a bespoke chat UI.
Argues chat won't replace everything, but it reshapes the UI landscape.

added by Adam Tomat • 25th Jun 2026

BlogUX Collective • Patrick Neeman

The T-shaped UX Professional is Giving Way to the Polymath Architect

AI is dissolving the classic "T-shaped" specialist model in UX toward generalist "polymath architects".

AI removes the cost penalty of working outside your specialty.
91% of designers now use AI weekly.
The scarce skill becomes directing AI and catching plausible-but-wrong output.

added by Adam Tomat • 23rd Jun 2026

BlogPolar

Building an LLM-safe design system (Polar's Orbit)

A typed component design system built to keep LLM-generated UI on-brand.

Typed components constrain what the LLM can produce.
"Safe" means on-brand by construction, not by review.
Complements agent design skills.

added by Adam Tomat • 18th Jun 2026

DocsAnthropic

Claude Code and Claude Design Now Sync Both Ways

Resolved via search — verify before relying on it. A `/design-sync` command that lets Claude Design and Claude Code pull each other's work back and forth, instead of a one-way design-to-code handoff.

Run `/design-sync` to pull your real design system into Claude Code, so components built there start from your actual buttons and tokens, not a screenshot.
Push what you build in Claude Code back to Claude Design and it stays editable on the canvas.
In beta at claude.ai/design and the desktop app, included on Pro, Max, Team, and Enterprise plans.

added by Adam Tomat • 17th Jun 2026

DocsAnthropic

frontend-design SKILL.md (anthropics/skills)

Anthropic's own skill for producing distinctive, intentional frontend design — not generic templates.

A two-pass process: draft a compact design plan, then self-critique for genericness.
Names three "AI-default" clichés to avoid.
Build with real content, and spend "boldness" on one signature element.

added by Adam Tomat • 16th Jun 2026