π‘ TJ highlighted the Design System Kit tool within Figma Console MCP as an alternative to figmalint for enterprise users blocked from personal API keys. It mirrors figmalint's design system validation capabilities through the MCP interface, usable in Cursor or Claude Code workflows.
π‘ TJ shared Figma Console MCP tool links as alternatives to figmalint for users without personal API key access. The figma_lint_design tool provides the same Figma linting functionality through the MCP, making it accessible via Claude Code, VS Code, or Cursor without needing direct API keys.
π₯ Brad's Claude artifact walking through his specific concoction of bfw-process, rules, AGENTS.md, skills, Eddie Brain MCP features, and validation checks β the reference document for keeping AI on design system rails.
π₯ Brad demos generating a footer component with and without design system guardrails to illustrate the difference β part of the new course chapter tackling how to actually keep AI on the rails of your design system.
π₯ Brad details his full setup for keeping AI agents on the rails of the Eddie design system: bfw-process phases, AGENTS.md, Claude skills, Eddie Brain MCP validation, and custom rules. The most complete look yet at the BFW AI guardrails stack.
π‘ Brad shared a Claude artifact that documents his full concoction for AI-guardrailed design system work: bfw-process, AGENTS.md rules, skills, Eddie Brain MCP, and validation checks β a living reference for how BFW keeps AI agents on the Eddie rails.
π€ Lauren shared this Wharton research paper on 'cognitive surrender' β adopting AI outputs with minimal scrutiny, overriding both intuition and deliberation. Sparked a thread about maintaining human agency in AI-assisted design system work and whether separate GitHub accounts for AI agents make sense.
π¬ Academic paper coining 'cognitive surrender' β the passive, uncritical acceptance of AI-generated outputs β as distinct from deliberate cognitive offloading. Proposes a Tri-System Theory adding AI as a third cognitive layer alongside fast and slow human thinking. Sparked a philosophical discussion in the channel about defining human vs. AI roles in design system work to avoid blind reliance on AI judgment.
π₯ Brad invites course community members to share real-world stories of operationalizing AI & Design Systems at their organizations β navigating buy-in, security hurdles, governance, pilots, and scale β directly in this FigJam board as input for Chapter 6.
π‘ Shared in a thread on the southleft.com redesign launch β a behind-the-scenes Substack post from Southleft on the why's and how's of using AI while keeping the design system in control of final decisions.
π₯ The new Southleft.com site went live, featuring fun Easter eggs and an interactive color-theme picker in the header. Thread replies showed community sharing favorite themes, and a behind-the-scenes Substack post was linked covering how AI and design system thinking shaped the build.
π₯ The new Southleft.com website launched today, featuring fun Easter eggs including an interactive element in the header. A follow-up shared screenshots of the site's 'Coastal Sea Breeze' theme. Southleft is the design systems consultancy closely connected to this community's AI & design systems work.
π₯ Brad Frost walks through the end-to-end process of redesigning and rebuilding bradfrost.com using the Eddie design system, showing how AI can be applied across the entire workflow β from design to implementation.
π₯ TJ Pitre puts five AI prototyping tools β Figma Make, Bolt, v0, Claude Design, and Lovable β through the same assignment and design system to see how they each perform. Community members are excited and asking great follow-up questions about code quality and custom design system support.
β Brad shared dembrandt.com, a tool with a /extract-design command that pulls real design tokens β colors, typography, spacing, components β from any live URL directly into an AI agent, with W3C output. He was curious if anyone in the community had tried it yet.
π₯ Brad announced jam session #10 is live, where TJ digs into Component Contracts and the crew covers the completion of Chapter 3. Part of a packed Friday update that also confirmed Chapter 3 is fully published and Chapter 4 recording is underway.
π‘ Brad announced that a Chapter 3 demo video β the walkthrough of installing and using the DS Inspection Kit β had been stuck in draft mode and is now live. Essential viewing for anyone following along with the Chapter 3 inspection workflow.
π‘ A UX Collective article introducing 'hypertokens' β a proposed semantic layer sitting between design tokens and components, designed to give AI agents richer contextual meaning when working with design systems. Brad shared it without commentary, letting the concept speak for itself.
π₯ Brad shared Spektral, a specification-first design system tool from longtime collaborator Kevin Coyle. The idea: with a strong spec in place, you can generate your design system in any format you need β Brad sees this as part of a broader industry move toward implementation-agnostic design system specifications. Kevin himself replied confirming his passionate evangelism for the approach.
π₯ Brad Frost announced the completion of Chapter 3 of the AI & Design Systems course β 64 lessons and over four hours of content on using AI to improve design system quality through a multi-point inspection framework covering completeness, soundness, synchronization, extensibility, and AI-readiness. Live demos on the Eddie design system took it from a score of 70 to 83 after addressing flagged issues.
π‘ Community member Doug Scamahorn shared a playground repo for exploring AI & Design Systems course concepts, built on the Figma Simple Design System. It integrates the Figma Console MCP server and desktop bridge plugin for seamless code-to-design and design-to-code workflows.
π₯ Doug Scamahorn shared a live demo of the figma-console MCP server and desktop bridge plugin combo in action, showing seamless code-to-design and design-to-code round-tripping. He built a playground on top of the Figma Simple Design System to explore course concepts.
π₯ The AI Design Systems Inspection Kit is a set of Claude skills for running an automated multi-point audit of your design system. Clone the repo into ~/.claude/skills/ds-inspection, run /ds-inspection, and get a scored, prioritized report covering completeness, soundness, synchronization, extensibility, and AI-readiness.
π₯ Brad shared a new course video walking through the Design System Multi-Point Inspection Kit β a set of Claude skills that audits your design system against the chapter's 10-station framework and returns a prioritized work order. The Eddie Design System scored 70/100, and Brad is now using the results to guide the final chapter demos.
π₯ Brad Frost shared an overview video walking through the five design system inspection qualities introduced in Chapter 3 of the AI & Design Systems Course: Complete (coverage & gaps), Sound (best practices, accessibility, shared language, testing), Synchronized (orchestration), Extensible (governance, feedback & adoption), and AI-Ready (machine-readable docs, agent access). The framework is designed to be durable as the tooling landscape evolves.
π₯ Brad Frost introduced TJ Pitre's concept of the 'Check Engine Light for your design system' β a metaphor for knowing when your system needs attention before it breaks down. This lesson grounds Chapter 3 of the AI & Design Systems Course, which structures design system improvement as a multi-point inspection across five qualities: Complete, Sound, Synchronized, Extensible, and AI-Ready.
π₯ Dan Donald built Sparebrain using Claude Fable β in under a day, from a detailed spec, with no special prompt magic. It's a pre-research hypothesis tool: you create personas, projects, and studies to explore product copy and pitches before doing full research. Currently in very early days but open to try.
π₯ Brad Frost shared this June 18th AI & Design Systems jam session covering Claude Design, agentic design systems, and more β recommended as holiday weekend viewing to help community members stay current on the breakneck AI and design systems zeitgeist.
π₯ Brad Frost shared TJ Pitre's July 2nd jam session covering post-Figma Config news, Figma Motion, v0, and AI-ready design systems β a roundup of the week's biggest developments recommended as holiday weekend viewing.
π‘ Brad Frost's bare-bones Storybook from his Atomic Design course, shared as a beginner-friendly starting point for AI & Design Systems students who want to clone and practice with Storybook and Claude. A simpler starting repo than a full production design system.
π‘ TJ Pitre's Altitude design system is available as a demo and practice resource for AI & Design Systems course students. Brad Frost recommended it as a great starting point for cloning a real Storybook locally and practicing component-to-code workflows with Claude.
π‘ Brad Frost shared the Storybook Showcase gallery as a resource for AI & Design Systems course students looking for real-world Storybook implementations to clone locally and practice with Claude. Cloning an existing live Storybook is a great way to get hands-on with the Storybook MCP without building from scratch.
π‘ Brad Frost recommended Storybook's official tutorials to a community member learning how to connect Storybook to Claude for component-to-code workflows. Covers getting Storybook set up from scratch β a prerequisite before cloning real-world Storybook repos to practice with Claude.
β Robin Di Capua ran into a Cloudflare UI navigation issue while setting up the Company Docs MCP tool by Southleft. The thread resolved quickly β the Account ID moved to Build > Compute > Workers & Pages after a Cloudflare redesign, and TJ Pitre noted he needs to update the README to reflect the new location.
π‘ Brad Frost shared this course lesson as a reference for tying product and system work context together through Git and Gitflow workflows, pointing community members to it as a foundation for AI-augmented design system documentation practices.
π‘ Rafael Matos da Silva shared a Claude Skill he built for assessing Figma/code parity across design systems. He's been fine-tuning it across multiple design systems and shared it with the AI & Design Systems course community to try out.
π‘ Brad Frost referenced the Waypoint design system as a model for practicing AI-ready design system workflows in personal environments outside company walls, recommending it to designers who want to experiment with Claude Code and design system tooling before moving to enterprise setups.
π₯ Brad announced that TJ Pitre's new lessons on bolstering documentation with company docs and Figma MCP tools are now live on the AI & Design Systems course platform. Community members were tagged to check them out.
π₯ A demo video of TJ Pitre's company docs MCP tool showing its visual output capabilities. Shared with fire reactions from the community, with the caption noting it handles the 'visual polish' side of documentation generation.
π€ TJ Pitre (Southleft) shares opinionated thoughts on why the term 'agentic design systems' makes them uncomfortable, despite using agents constantly in practice. A nuanced take on terminology and what it actually means to integrate AI agents into design system workflows.
β Brad asked if anyone in the community has tried the Buddy Figma plugin by Anima, which uses Claude to generate designs, HTML, and code from natural language directly in Figma.
π‘ Brad Frost shared Web Awesome's newly released agent skills documentation for their component library, noting it's valuable even outside of Web Awesome itself β the approach to bundling AI agent skills for a component library is generalizable inspiration for any design system team building similar workflows.
π₯ Community member Mike Feldman shared this fresh Claude AI post on X, noting 'Just dropped.' The share landed during the team's live jam session where they used Claude Design to ingest the Eddie design system, making the timing relevant to active #ai-and-design-systems course discussions.
β Robin Di Capua shared a LinkedIn post about Claude Design's new on-brand capability, asking the community if it's the real deal or still limited. The update enables Claude Design to maintain brand consistency across projects using design system imports, with org-wide admin controls to lock brand standards across all work.
π₯ Brad Frost enthusiastically shared Christine Vallaure's deep-dive into A2UI (Agent-to-User Interface), a protocol enabling AI agents to dynamically generate context-aware UI. The thread drew excited reactions β Christine herself replied 'I am OBSESSED!' and community member Jennifer Robbins praised it as guiding seasoned designers through new territory.
π‘ Polar shared their blog post about Orbit, an LLM-safe design system, as context in a thread discussing Emil Widlund's CSS-in-JS post. Relevant to ongoing conversations about how design system architecture choices affect LLM comprehension and code generation quality.
π₯ A community member shared the Alert component from the Waypoint Design System, noting they're working on it externally and bringing it to their team at ADP. Brad responded enthusiastically, calling it badass and glad it's proving useful in real-world adoption.
π¬ Juan Di Nella shared an X post by Emil Widlund showcasing a CSS-in-JS approach. Brad noted it's interesting to see momentum moving back towards CSS-in-JS, bucking the recent industry trend away from it.
π‘ Jeremy Richardson shared this Syntax FM episode while discussing when AI is β and isn't β the right tool. He found it helpful for thinking through using deterministic codemods instead of AI for migration tasks. The thread sparked broader discussion about differentiating AI from pure automation, with Ian Frost noting the same lesson applies to design system component API changes.
π‘ Shared in a deep thread on DTCG compliance and dropping Tokens Studio from the token pipeline. Rafael Sepeda linked this to show that Figma's extended collections can now handle white label scenarios and are supported by Supernova β relevant context for teams evaluating a native Figma variables + Supernova-only pipeline.
π‘ Brad Frost shared this article by Emily Campbell, noting it articulates something he's been working on for months. It explores the layered nature of AI experience β a useful conceptual frame for thinking about how AI integrates into products and workflows.
π₯ TJ Pitre announced he's joining Figma to focus on AI and design systems, sparking an outpouring of excitement from the community (17 hearts, 13 tada reactions). TJ confirmed the AI & Design Systems course continues as business as usual.
π‘ Craig Mertan shared a primitive color builder tool featuring improved transitions between color swatches and built-in color harmony support β a useful utility for design systems teams working on color scale generation.
π₯ Jesse Gardner shared an instructional guide covering the full spectrum of Web Component styling approaches β from locked CSS variable overrides to fully open light DOM β helping teams decide the right level of flexibility per component. The thread sparked a lively discussion with Brad and Ian Frost about Shadow DOM pain points, Lit framework wrappers, and how reducing component complexity also reduces LLM context costs.
π‘ Brad Frost shared this LukeW article, noting it articulates something he's been working on for months. The piece argues that design tools themselves are becoming the new design deliverable β shifting the conversation from static exports to living, tool-driven artifacts.
π₯ Brad Frost shared this LinkedIn post with a pointed π β a 12-minute video covering Claude Code tips aimed specifically at designers. Community reaction noted the honest framing that 'Claude isn't good at design... yet.'
π¬ Nathan Curtis's post on treating components as structured data was referenced in a thread comparing YAML vs JSON approaches for component definition β with Nathan emphasizing that the real goal is implementation-agnostic source of truth that spans files, content types, and maximizes regeneratability and verifiability.
β PJ Onori's Design System Documentation Spec was surfaced as an alternative structured approach to capturing component definitions. Nathan Curtis provided context on how it differs from his Components as Data thinking β noting that the source of truth must be independent of Figma and Markdown files are too limited for robust design intent storage.
π¬ TJ Pitre's Southleft newsletter roundup of current design systems and AI reading, which sparked a community thread comparing the Design System Documentation Spec (DSDS) to Nathan Curtis's 'Components as Data' approach β with Nathan himself weighing in on why source-of-truth independence from Figma and Markdown limitations are driving his decisions.
π₯ Brad shared an Instagram reel with a single-word caption β 'ACCURATE.' β a relatable, shareable moment that resonated with the channel's AI and design systems vibe.
π₯ Brad Frost flagged GitHub's new 'app' repo β home for releases and issues for the GitHub Copilot app. TJ Pitre noted that GitHub reserving a top-level repo named just 'app' signals how seriously they're treating it.
π₯ Brad Frost announced new Storybook MCP content in the AI & Design Systems course, including this intro lesson by TJ Pitre. The channel lit up with 10 tada reactions and 9 raised hands.
π₯ Brad Frost added a live Storybook MCP demo from Storybook founder Dominic Nguyen to the AI & Design Systems course β a great companion to the intro lesson showing the tool in real action.
π‘ Jesse Gardner (plasticmind) built an interactive tutorial explaining how the ElementInternals browser API works with web components, with a strong emphasis on accessibility. It covers how custom elements can participate in forms, handle validation, and manage accessibility relationships the same way native HTML elements do β essential reading for design system engineers building with Lit or other web component approaches.
π₯ TJ Pitre released a public repo of 22 Figma Console MCP tools converted into skills that run on the native Figma MCP, covering token export/import, accessibility audits, design-system linting, and code/design parity. Compatible with Claude Code, Claude Desktop, and other platforms.
π₯ TJ Pitre shared a LinkedIn video walkthrough demonstrating how to install and run Figma Console MCP skills with the native Figma MCP, showing tools like token sync, accessibility audits, and design-system linting in action.
π‘ Brad shared the Claude Code Routines docs in the middle of a lively thread about building automated personal digest workflows. The thread had Rafael building a NotebookLM pipeline, Robin building a daily-direction system, and Jenelle asking how non-devs can pull this off β Brad dropped this as the low-friction answer for Claude users who want scheduled tasks without scripting.
π₯ A new AI & Design Systems course lesson clarifying the difference between Figma MCP Connectors, Plugins, and Skills β key conceptual distinctions for anyone building or using Figma-integrated AI workflows.
π₯ Brad shared a new AI & Design Systems course lesson walking through how to install the Figma MCP connector and plugin in Claude Desktop β a direct setup guide for getting the Figma MCP integration up and running.
π₯ Chapter 4 course lesson demonstrating how to use Figma MCP to generate a design-system-powered login page β a practical end-to-end example of AI-assisted UI generation directly from Figma's design system.
π‘ Dan Donald published a Medium piece explaining the background behind his Kickstart tool β what it is and how he built it β aimed at helping less technical folks get started with Claude Code. A practical on-ramp for non-developers wanting to work with AI coding tools.
π₯ A community member built an AI Inclusion Skill for their Waypoint design system as a POC β it runs pre-deployment international and cultural inclusion reviews, covering RTL/LTR layout, text expansion budgets, ICU plurals, color symbolism, gesture iconography, and translation-readiness. Dan Donald flagged his gotrino-assistant plugin as a related resource with overlapping capabilities.
π‘ Dan Donald shared this Figma site focused on UI patterns for AI products, noting it might be of interest to the community. Brad Frost chimed in that his upcoming interview with Dan Saffer (another voice in the AI UI space) should be dropping in the next week or two.
π‘ Brad Frost shared this 2026 report exploring how designers are evolving their tools, craft, and teams with AI β a useful pulse-check on the current state of AI adoption across the design profession.
π₯ The third and final installment of a three-part series on the future of design systems, published on zeroheight β this piece tackles the 'diagnostic layer,' covering how to close the gap between design and code and where the source of truth should actually live. Got strong positive reception (5 hearts, 4 thumbs up) with the thread devolving into a fun exchange about chaotic publishing schedules.
π‘ Brad shared a LinkedIn post asking community members to boost or leave testimonials for the AI & Design Systems course ahead of a price increase the following Tuesday. A lighthearted ask to help spread the word to colleagues who might benefit.
π₯ TJ Pitre shares a new Southleft piece making the case for declarative over designed documentation when bringing AI into design systems. Based on hands-on testing, the article argues that semantic intent declared in component descriptions (plain Markdown) gives AI far better, more reliable context than trying to interpret visually-structured Figma frames.
π₯ Davy Fung shared this video from @miggi on X, drawing enthusiastic reactions from the community. Content appears to be a demo or exploration relevant to AI and design systems.
π‘ Shared during the AI & Design Systems jam session recap by Ian Frost. Superpowers is a skills framework plugin for Claude Code by Jesse Vincent, providing structured, composable workflows for software development β including TDD, debugging, brainstorming, and subagent-driven development with built-in code review.
π€ Shared during the AI & Design Systems jam session recap. The 2026 WebAIM Million found 95.9% of home pages still have detectable WCAG 2 failures β reversing years of gradual improvement. A sobering baseline as AI-generated UI becomes more common in production.
β Brad Frost asked if anyone has used Buddy, a Figma community plugin that acts as an AI design agent β generating, extending, and iterating UI designs directly in the Figma canvas using Claude.
π‘ Brad shared Chrome's Modern Web Guidance docs β a practical reference for building on the web platform using up-to-date best practices and APIs.
π₯ TJ shared a new demo building a design system from scratch with AI starting from a completely empty folder β no existing codebase scaffolding. The key takeaway: getting context set up before the AI builds anything is what actually matters.
π₯ Figma announced their new Design Agent, sparking excitement in the community. Thread discussion explored whether Figma's native AI integration would outperform external workflows like Claude Code + MCP β with a resounding 'Yes, 100%' from the group.
π‘ Dan Donald shared Kickstart, a Claude Code starter template aimed at less technical users β fork it, run /onboarding, and Claude interviews you about your project, writes a tailored CLAUDE.md, and sets up session memory and weekly self-update skills. Dan was looking for community testers.
π¬ Brad shared this post from Mark Erikson (Redux maintainer) in the #ai-and-design-systems thread on grief and AI β describing it as an honest personal account of one developer's evolving fears, opinions, and mental journey with AI. A candid, grounded perspective from someone deep in the JS ecosystem.
π¬ Brad shared his blog post exploring the emotional and philosophical dimensions of navigating AI in one's work and life, inviting the community to reflect on their own journeys. The thread sparked a follow-up share of another developer's honest personal account of fears and mental shifts around AI adoption.
π‘ Dan Donald (hereinthehive) shared this follow-up post to encourage less technical folks to get hands-on with making agents β breaking down the anatomy of an agent in approachable terms for designers and non-engineers exploring agentic workflows.
π‘ Lauren shared this self-paced online course by Sara Soueidan aimed at web designers and developers who want to build more accessible UIs. Though not AI-specific, the thread noted it's highly regarded for building a solid accessibility foundation β a skill increasingly relevant as AI-generated UI code often misses accessibility basics.
π‘ Rafael shared this deep-dive on Pi, the minimal agent runtime powering OpenClaw, in the thread where Brad asked if anyone had tried Pi. Good technical explainer on how Pi works as a lightweight agent framework β relevant context for anyone thinking about agentic tooling in design system workflows.
β Brad is asking if anyone has tried Pi, a minimal terminal-based AI coding agent that runs locally with no SaaS backend, supports 15+ model providers, and is MIT-licensed. The thinking_face reaction suggests the community is curious but hasn't weighed in yet.
π₯ Chapter 2 of the AI & Design Systems Course is officially live, covering AI tool installation (Cursor, Claude, Codex), API key configuration, and git/GitHub setup. Community members are finding it particularly valuable β especially the Git/GitHub section β even for experienced developers.
π‘ A YouTube video covering major new Notion feature announcements was shared for Notion power users in the channel. No additional context was provided beyond the link.
π₯ TJ Pitre shared a LinkedIn post on enterprise design systems and the check engine light metaphor β the idea that design systems surface deeper organizational problems rather than cause them. The post directly connects to Chapter 1 of the AI & Design Systems course and received an enthusiastic response from the group.
π‘ TJ Pitre shared this follow-up Substack piece to the Context-Based Design Systems Revisited article. It goes deeper on how context cascades through a design system and what fundamentally changes when AI agents enter the picture β extending the CBDS framework into more concrete, agent-aware territory.
π₯ Christine Vallaure called out this full data report from Zeroheight's State of AI in Design Systems research β the detailed findings companion to their overview page. The report contains survey data on how design system teams are adopting AI, offering both validation and directional signals for where the field is heading.
π‘ zeroheight published a data-rich report on AI usage in design systems, offering concrete findings on where teams are investing and what's working. Brad Frost flagged it for the next course jam session to dig into the findings together.
π₯ The companion demo site to Brad Frost's new Git course module β a living Git cheat sheet built using the exact concepts and workflows covered in the lessons, making it a practical reference and a meta-demonstration of AI and Git working together.
π₯ Brad Frost announced 20 new AI & Design Systems course lessons covering Git and GitHub fundamentals through to AI-generated commits, branching, and supercharging GitHub Issues with AI. The module frames Git as the connective tissue β 'AI as mortar' β that binds AI workflows to real design systems work.
π‘ Dan Donald makes the case for designing your own agentic workflows with Claude Code rather than defaulting to off-the-shelf tools. Community replies resonated strongly β Ian Frost and others noted how important (and uncomfortable) it is to break the 'what plugin should I use?' mindset and start building custom solutions.
π‘ TJ Pitre shared that Southleft is expanding their contractor pool, with openings primarily in design systems and context engineering. Relevant to the AI & design systems community as Southleft continues to grow their work in this space.
π‘ TJ Pitre referenced this Claude docs page in a detailed breakdown of Figma CLI vs Figma Console MCP token costs β explaining that Claude Code now defers MCP tool loading by default, pulling in only the 3-5 tools needed per query rather than injecting full schema upfront. Key context for understanding why MCP token overhead is lower than commonly assumed in modern AI coding clients.
π‘ Robin Di Capua shared this Figma Console MCP docs page comparing the official Figma MCP with the Figma Console MCP, noting it helped clarify the key differences. The discussion touched on token usage efficiency and whether batching via Figma Console MCP changes the trade-offs.
π‘ Keith shared Roger Wong's blog as a helpful resource for navigating today's AI landscape, in response to Brad Frost's channel-wide question asking for recommended blogs, newsletters, and websites for the design systems community.
π₯ TJ Pitre from Southleft revisits Context-Based Design Systems, outlining a workflow where designers use AI to produce the first coded version of a component themselves, then a context engineer takes over to enforce coding standards, write tests, validate accessibility, and ship. Brad shared it while actively recording Git/GitHub videos to help non-developers enter this workflow β the community noted the article's excellent timing alongside the recent DESIGN.md jam session.
π₯ The May 7 weekly jam session video is now posted, covering the DESIGN.md trend, Claude Design, and related topics. Part of the ongoing AI & Design Systems course jam session series.
π₯ Brad, Ian, and TJ announced that Chapter 1 of the AI & Design Systems Course is officially live, covering the broad arc of technology, design systems history and value, AI history and value, and tooling orientation. The community responded with excitement β 5 tada reactions and counting.
π‘ Brad and Ian Frost joined Chris Coyier and Dave Rupert on ShopTalk Show to discuss AI & Design Systems, exploring where developers are in their thinking about AI in development generally. Brad noted there are some good nuggets in the conversation.
π₯ Brad Frost joined Donnie's Wireframe Live show for a wide-ranging conversation about AI & design systems, with Brad noting they got into some "spicy" territory. The community responded enthusiastically, racking up hearts and robot reactions.
π‘ TJ Pitre shares an article noting that most enterprise design system teams still lack AI tool access β and the design hygiene practices that prepare a system for AI turn out to be things teams should have been doing all along. Directly relevant to the course jam session on getting Figma files ready for AI tools they don't have yet.
π₯ Brad shared Sarah Drasner's new illustration explaining context windows, praising her background as a medical illustrator as uniquely suited to making visual explainers for complex AI concepts.
π¬ A community member flagged a backlash thread against non-technical people shipping to production thanks to AI, sparking a rich discussion about guardrails, empowerment, and what it actually means to ship. Ian Frost and others argued for empowerment with proper infrastructure β a theme the course will dig into directly.
π₯ The live demo website Brad built using a chain of off-the-shelf AI tools: Claude generated a medieval poem about Atomic Design, ChatGPT illustrated it, Suno turned it into a rock opera, Claude Code with Figma Console MCP designed and then coded the website β with a second design token theme for the space ballet sequel added for good measure.
π₯ The unedited full-length version of the LLMs in Action session, added to the course Appendix with more context and commentary throughout the AI tool chaining process. Recommended for those who want the complete journey.
π₯ Brad released the edited LLMs in Action course lesson, a wild rabbit hole session chaining Claude Chat, ChatGPT image gen, Suno, DaVinci Video Generator, and Claude Code with Figma Console MCP to build a medieval rock opera website β culminating in the central course question: IS IT GOOD? Community greeted it with delight.
π‘ Surfaced by Lauren Basser while asking about AI-ready documentation before a codebase exists. Nathan Curtis's specs-cli generates Figma component specs at 1 second per component with 99%+ compression β Nathan himself joined the thread to discuss fine vs. coarse artifact sources for AI-ready design systems.
π‘ Brad released a 10-minute Brief History of AI course lesson (that took 2.5 hours and months of research to make), contextualizing the current AI moment for design systems practitioners. The final piece needed to complete Chapter 1 of the course.
π€ Brad Frost shared a GitHub template for cloning any website with one command using AI coding agents β accompanied by a deadpan π« reaction, suggesting equal parts fascination and existential unease at how accessible this kind of capability has become.
π‘ Vincent Schlothauer shared Ian Guisard's article exploring the concept of Component.md β on a wavelength with the ongoing 'is Figma dead?' discourse and how the role of traditional design tooling is shifting as AI takes on more of the assembly work.
π¬ TJ Pitre (Southleft) wrote an op-ed after actually spending time with Claude Design, arguing it's not a design systems tool β and that's fine. The thread got spicy when Brad Frost coined 'Design System Cosplay' and 'Uncanny Valley UI' to describe AI that approximates your design system without actually using it β both terms landed hard with the community.
π₯ Jem Gold shared her X post promoting 'Design Is How It Tastes' and asked the community for RTs.
π₯ Jem Gold makes a compelling case that design specs focused on object-language (tokens, radii, spacing) strip out the felt intention of design, proposing 'encounter-language' β sensory, atmospheric prompts fed to image models β as the generative semantic layer above the system. A genuinely fresh angle on where design systems thinking meets generative AI.
π₯ Lou Manning (Design Systems Director at ADP) shared an overview of Waypoint, his evolving AI-ready design system exploration β with background on how it started and how it's matured into a potential future system. He's actively building a machine-readable Rules skill and working on Phase 5: closing the loop by updating Figma assets via the Figma Console MCP. TJ Pitre responded enthusiastically noting Console MCP was the unlock Lou needed.
π₯ Brad Frost shared this essay exploring what human qualities remain irreplaceable as AI automates production β arguing that what becomes scarce is the human involvement that makes products feel made for someone, by someone. The thread sparked personal reflections on creativity and the shift toward inherently human value, with Luana Cruz connecting it to her own growth in the community.
π‘ Ian Frost shared a LinkedIn post about AI and design systems work β his first LinkedIn post in a while, inspired by the community's activity. The community responded warmly.
π‘ Kaelig's Bluesky cross-post of his article on building design system components with AI agent teams, shared alongside LinkedIn and X posts. The article covers his Intuit-era work on using AI agent teams to automate design system component work.
π‘ Kaelig (formerly of Intuit) shared his LinkedIn post announcing the full publication of his article on building design system components with AI agent teams β posted alongside a Bluesky and X cross-post, asking community members to repost as he searches for his next role.
π‘ Kaelig's X cross-post of his article on building design system components with AI agent teams, shared alongside LinkedIn and Bluesky posts. The article covers his Intuit-era work on using AI agent teams to automate design system component work.
π₯ Kaelig published this long-anticipated article on using AI agent teams to build design system components, shared shortly after leaving Intuit. The community greeted it with enthusiastic raised-hands reactions.
π₯ Brad Frost shared his blog post about 'mouth coding' β verbally directing AI to redesign and rebuild his wife's counseling practice website in under an hour. Thread replies showed others were immediately inspired to do the same for their own projects.
π‘ Brad Frost shared Luke Wroblewski's article on how AI tools are reshaping the designer-developer relationship, noting it aligns with many concepts covered in the AI & Design Systems course and emphasizes the opportunity for better human collaboration rather than just tool replacement.
π‘ Michael Helmbrecht shared CDF (Component Description Format), a declarative, design-system-agnostic spec for documenting UI components in a way LLMs can read and author. Covers component properties, states, tokens, and accessibility; validated against five real-world design systems including Radix Primitives, Primer, and Material 3.
π¬ Lauren Slowik shared Anthropic's Claude Design announcement, triggering one of the most active threads in the channel (19 replies). Community reactions ranged from exhaustion and excitement to concern about Figma's position, with discussion of token costs, design system import support, and whether this signals Anthropic competing directly with Figma Make.
π‘ Brad, Ian, and TJ describe their 'Progressive JPEG' methodology for creating the AI & Design Systems course β building iteratively with increasing fidelity rather than front-loading all production quality, mirroring how they teach AI-assisted development itself.
π₯ Brad Frost shared a course update newsletter announcing the AI & Design Systems course has now accumulated 14.5 hours of video content, along with a description of the team's 'Progressive JPEG' approach to building the course iteratively.
β Robin Di Capua raised this Figma CLI tool (by Sil from Into Design Systems) as a point of comparison against Figma Console MCP. Thread discussion clarified that the CLI assumes a shadcn/Tailwind stack while the MCP is stack-agnostic β making the right choice depend on your existing setup and use case.
π‘ Dan Donald released AI Radar, a Claude Code plugin that runs targeted web scans based on your project context to surface relevant AI developments β addressing the community's widespread feeling of overwhelm keeping up with the pace of change in AI and design systems.
π‘ Brad Frost referenced this foundational post in a thread challenging the notion of a singular 'source of truth' β arguing that as AI systems grow more sophisticated, no single tool (including Figma) can hold that role; the entire ecosystem of code, docs, agents, and design is the source of truth.
π‘ Brad Frost shared this Storybook MCP demo video featuring a conversation with Dominic Nyugen, referenced in a broader discussion about source of truth and whether Storybook adds meaningful value in AI-driven design system workflows alongside direct codebase access.
β Zdenek Zenger raised troubleshooting issues running Figma Console MCP in Windsurf vs. Claude Desktop. A 9-reply thread surfaced a key conflict: having both the NPX version and the claude.ai connector registered simultaneously causes failures β removing the cloud connector from claude.ai settings resolved the issue for multiple people.
π‘ Shared by Dimitri during the AI & Design Systems jam session and later recapped by Ian Frost. A YouTube video titled 'AI Reacts to Anxiety' that came up during the broader conversation about ethics, humanity, and AI's role in design.
π‘ Official Claude Code docs for hooks β the lifecycle events (SessionStart, PreToolUse, PostToolUse, etc.) you can wire into your own workflows. Shared in Ian's jam session recap as a jumping-off point for folks wanting to customize Claude Code for design system tasks.
π‘ Dan Donald's evolving gist on ethics and inclusive design skills β a living 'Design System Community Charter' he links from his core context file when working with LLMs. Shared in Ian's recap of the jam session as a practical example of weaving ethical guardrails into AI-assisted design system work.
π‘ Shared by Ian Frost in the recap from the AI & Design Systems jam session with Dan Donald. Designpowers is an open-source workflow of 10 specialist agents that collaborate to run an inclusive design process β from discovery through retrospective β while the user stays in the director's seat.
π€ Google Stitch's overview page for DESIGN.md files β the new context-file pattern getting hyped across design Twitter. Shared alongside an experiment testing whether they actually change AI output. Thread discussion pushed back skeptically: Dan Donald and Vincent Schlothauer noted it's 'just another markdown file' and like any context, its relevance degrades over longer sessions.
π¬ TJ Pitre ran two Claude Code sessions against the same Figma file using Figma Console MCP β one with a DESIGN.md, one without β to actually test whether the new DESIGN.md pattern makes a difference. TLDR: they work, but they're just context files, same pattern as CLAUDE.md or .cursorrules. Thread discussion echoed the skepticism about hype around yet-another-markdown-file.
π‘ Dan Donald shared an update to his Claude Code plugin, now called gotrino-assistant, which adds a triage skill for task evaluation and prioritization. It builds on ethics and inclusion principles to help with project management, and he's considering porting over more of his daily-use tools.
π€ Brad shared Mike Davidson's article on career advice for designers navigating 2026's AI landscape, arguing that 'the assembly layer is going away' and that designers must shift from specification and assembly work toward orchestrating AI tools and building functional products. The piece offers concrete advice on building AI-fluent portfolios and embracing a conducting role over a creating role.
π₯ TJ Pitre announced new accessibility tooling in Figma Console MCP that runs WCAG 2.2 checks on both design (contrast ratios, target sizes, focus indicators, heading hierarchy) and code (axe-core scanning for aria issues, semantic problems, and labeling gaps). The tools can be chained together with a parity check to keep design and code aligned β positioned as a deterministic safety net for catching easy-to-miss issues, not a replacement for human accessibility expertise.
π₯ Brad shared his LinkedIn post announcing an exclusive first-look demo of the new Storybook MCP with founder Dominic Nguyen, highlighting how AI-generated UI running through design system testing and tooling is a bullseye for the AI & Design Systems course thesis.
π₯ Brad shared his wide-ranging conversation with Dominic Nguyen featuring a live demo of the brand-new Storybook MCP. The community had been asking to see the MCP's features in action, and this video delivers exactly that.
π₯ Brad hosted a live Q&A with Dom from Storybook to dig into their new MCP, with the community asking about features like bringing design systems from Figma into Storybook and seeing everything in a single style guide view. The team demoed the MCP live during the session β this page is the official AI overview for Storybook.
π₯ TJ Pitre demos building a full 18-slide Figma Slides presentation template β with themed colors, typography, and everything bound to Brad and Ian Frost's design tokens β using Figma Console MCP in just 20 minutes and 3 prompts. Thread replies echoed similar wins, with one member using Claude to generate portfolio case studies from a markdown-defined template.
π₯ Christine Vallaure shares a designer's perspective on the agentic shift in design systems, arguing it's as much a design opportunity as a developer one. Inspired by a Storybook demo session, the post explores what good Figma file setup looks like to support agentic workflows. Enthusiastic community response with fire and heart reactions.
π‘ Shared in a thread reply as a recommended read alongside the Elyse Holladay podcast episode. Called 'a gem' by a community member.
π₯ Brad shared this podcast episode featuring Elyse Holladay discussing practical approaches to AI and design systems right now. A thread reply also highlighted a related zeroheight article as a gem.
π₯ Brad highlighted TJ Pitre's interview on the Design Systems Collective, where TJ discusses how AI enhances design systems work. Brad also linked his own earlier interview on the same publication, reinforcing the core 'Is it good?' principle of the AI & Design Systems course.
π‘ Brad referenced his own interview on the Design Systems Collective, noting it still captures his core thinking about design systems β especially the 'Is it good?' framework that underpins the AI & Design Systems course.
π₯ TJ Pitre shared a LinkedIn post about designers finally being able to own a component from Figma all the way to a pull request β not by learning to code, but by leveraging current AI tooling. Craig Mertan confirmed in the thread that he's doing exactly this with Figma Console MCP and it made the design process feel much more complete.
β Brad asked if anyone has built Figma Skills β a new community-driven marketplace for shareable AI skills in Figma. Thread discussion shows curiosity mixed with some skepticism about Figma crowdsourcing the community's work, plus comparisons to skills.sh.
π₯ Recording of the third AI & Design Systems jam session covering the evolving role of designers in an AI world, a generative design token system demo from Dimitri Otero, new tools like Paper.design and Pencil.dev, practical AI automation workflows, and FigmaLint.
π‘ CSS Tricks article exploring the CSS contrast-color() function, shared in the jam session recap for its relevance to design token systems and accessible color contrast automation.
π‘ FigmaLint plugin for cleaning up Figma files, featured in the AI & Design Systems jam session recap as a useful tool for maintaining design system quality.
π‘ Community FigJam file shared as part of the AI & Design Systems jam session resources, providing a framework for building agentic design systems.
π‘ Pencil.dev highlighted alongside Paper.design as a new visual collaboration tool that isn't Figma, discussed during the AI & Design Systems jam session as part of the 'design in the browser is back' conversation.
π‘ Jesse Basham shared a React-friendly Figma typography template in a thread where Christine Vallaure asked about best practices for typography styles and variables setup from a developer's perspective. Rafael Sepeda also shared his own typography accessibility guides inspired by Material and Apple patterns.
π‘ Survey collecting data on AI usage in design systems, building on the design systems report. Plans to publish findings as a report with practical guidance chapters. Community praised the quality of the questionnaire and one member mentioned recording a podcast about the topic.
β Jesse Basham asked if anyone uses Impeccable Style, a shareable set of design opinions for AI coding. Thread showed mixed reactions β one member uses it for their personal site, while others raised concerns about importing opinionated rules and suggested configurable setup skills instead.
π₯ Kaelig shared a teaser of building a complex Intuit design system React component using AI. Thread showed excitement, with praise for the Storybook MCP and mention of a fast follow-up to speed up reviewing changes in Storybook.
π₯ TJ Pitre shared a demo refocusing the conversation from Figma write-back hype to practical design-to-code workflows. Shows one-shot prompting of a fully specced Accordion component to production code using Figma Console MCP's annotation, token resolution, and interaction spec tooling. Generated significant thread discussion about token export workflows and using Console MCP with Lovable.
π₯ Rafael shared VS Code's v1.113 update highlighting new AI features including configurable thinking effort, nested subagents, and CLI agent capabilities with MCP server support β curious how these will speed up design system workflows.
π₯ Someone shared Agentation, a new tool they found intriguing. Thread discussion compared its point-to-edit approach to Figma Make, with the community seeing it as the future direction for frontend generation.
π₯ Brad shared registration for the upcoming live AI & Design Systems Jam Session, inviting community members to demo their work. Dimitri Otero replied expressing interest in presenting a computed design system that was only possible to build with AI.
π₯ TJ shares early access impressions of Figma's native MCP beta with new write capabilities, noting how it complements rather than competes with Figma Console MCP β the native tool handles execution while Console MCP focuses on bi-directional design systems management. Enthusiastic that the design systems community is influencing Figma's direction.
β A community member asked about the right balance of content to include in AI-facing MD files for design system components, referencing Uspec documentation they've been adding to Figma components alongside usage guidelines and org-specific context.
π₯ Jem shared their exploration of agents communicating through generative interfaces, recently refactored to use Claude Code's new Channels feature. An experimental look at how AI agents might use dynamic UI to interact with users.
π‘ Brad shared GitHub Next's Continuous AI project, noting it feels super relevant to design systems work. Explores how AI can be continuously integrated into development workflows.
π¬ Dan Donald shared this Here in the Hive piece inviting the community to reflect on where AI is heading and what futures we can (and can't) envision. A thought-provoking prompt for the group.
π¬ Brad shared Google's Design MD format spec from Stitch, a markdown-based standard for supplying agents with design context. The thread sparked discussion about the need for an open standard β similar to design tokens β for structuring component specs and design system context for AI workflows.
π‘ Brad recommended Anthropic's official course catalog as a foundational resource for learning Claude and AI skills. Shared during Into Design Systems conference day, where the thread turned into excited discussion about standout talks on design systems and AI integration.
π‘ Jonnie Lappen shared a talk from MCP Connect Day where Ruben Casas explores generative UI within MCP, pushing beyond traditional component-based approaches.
π₯ TJ Pitre announced major Story UI updates including a Voice Canvas for generating Storybook stories by voice (React), multi-framework support (Vue, Angular, Svelte, Web Components), and multi-provider LLM support (Claude, OpenAI, Gemini). Inspired by experiments from the AI and Design Systems course.
π₯ Brad shared a nearly three-hour video of himself redesigning bradfrost.com entirely through voice while painting a bathroom mural β no keyboard, no screen. He reflects on flow states, the creative infinite, and what it means when the mechanics of design and development no longer bottleneck creative output.
π‘ Kevin Coyle's Zeroheight blog post covers the central thesis of the AI & Design Systems course: design systems are critical infrastructure but cumbersome to manage, AI is powerful but messy, and combining the two produces better results faster while staying grounded in organizational reality.
π₯ Brad shared the second AI & Design Systems jam session covering voice-driven UI generation with Eddie, Claude + Figma MCP auditing, cross-platform design tokens, Penpot, and a course structure reveal.
π‘ Penpot was mentioned as an open-source design tool alternative during the second jam session, alongside discussion of OpenPencil and the Figma ecosystem.
π₯ TJ announced Cloud Mode for Figma Console MCP, enabling any web-based AI platform (Claude, Lovable, v0, Replit) to read and write Figma files via a pairing code β no Node.js or terminal required. Demo shows three platforms connected to three Figma files simultaneously.
π‘ Jem Gold's essay shared by Brad as a companion piece to their interview β exploring the implications of a world where everything just gets built with AI tooling.
π₯ Brad shared his drum and synth cover of RJD2's 'Ghostwriter' as a fun bonus link alongside the Jem Gold interview resources.
π‘ Another essay from Jem Gold's Superposition newsletter, shared by Brad as part of their conversation about creativity and AI's role in design and development.
π₯ Brad shared the first part of his wide-ranging conversation with Jem Gold covering AI, design systems, and creativity in the new agentic era. The community responded enthusiastically, with the post receiving raised hands and hearts.
π₯ Brad highlighted Jem's earlier demo showing it's possible to turn a hand-drawn sketch into a design system-powered website β a demo he says 'completely blew my mind' and still holds up as magical. He suggested someone should pick it up and run with it using today's more powerful tools.
π‘ Shared as a recommended watch alongside the Jem Gold interview β Jenny Wen's appearance on Lenny's Podcast discussing design systems and AI.
π‘ Brad shared Jem Gold's newsletter Superposition as a resource for the community, encouraging people to support Jem's work at the intersection of AI and design systems.
π‘ Brad referenced The Grid, Jem Gold's earlier AI-powered design startup from around 2015, as context for how long Jem has been working at the intersection of AI and design systems.
π₯ GitHub shares how the Figma MCP server can now generate design system specs, highlighting the growing ecosystem of MCP-powered design tooling.
π₯ Uber's design systems team published how they use Figma Console MCP to automate component spec creation β full specs (anatomy, API, tokens, accessibility) generated in Figma in minutes instead of weeks. Pipeline runs locally with no design data leaving the network.
π‘ TJ demonstrates the bi-directional nature of the Figma Console MCP, showing seamless connection and real-time interaction between AI models and Figma. Next step: comparing how different AI models interpret the same prompt.
π₯ LinkedIn post from Uber's team detailing how they built an agentic system using Figma Console MCP to automate design spec creation across iOS, Android, and Web platforms.
π¬ Brad Frost's blog post reflecting on the current state of AI from a designer's perspective β balancing opportunity with ethical responsibility.
π‘ The W3C's Ethical Web Principles document, referenced in discussions about how AI tooling should align with web standards and ethical practices.
π‘ Technical comparison of Skills-based vs MCP-based agent architectures, exploring different approaches to building AI agents that interact with design tools and codebases.
π‘ Supernova's February 2026 release introduces design system documentation directly within prototyping workflows, bridging the gap between documentation and hands-on design.
π‘ TJ shares an update on the Figma Console MCP, demonstrating new capabilities and answering community questions about the tool's direction.
π₯ Brad Frost explores the concept of real-time UI β interfaces that adapt and respond in real-time through AI, moving beyond static component rendering.
π₯ TJ demonstrates 'vibe coding' with design systems β using natural language to generate UI that stays within design system constraints, showing how MCP enables creative expression without breaking system rules.
π‘ Zeroheight explains how LLMs leverage MCP servers to read design system documentation, and provides practical guidance on writing prompts that work effectively with MCP-powered design systems.
π‘ Jem shares a hands-on experiment of spending a couple hours building with design systems and AI tooling, exploring practical workflows and noting areas where the tools shine vs. fall short.
π‘ Skills.sh directory listing for design token-related skills β reusable AI skills for working with design tokens in various contexts.
π‘ The official site for Southleft's Figma Console MCP β a tool that turns Figma into an operable console for AI agents, enabling read and write access across projects, variables, components, docs, and pages.
π‘ Documentation for the Figma Console MCP's component doc generation tool β automatically generates comprehensive component documentation from Figma designs.
π‘ Open source MCP server for design systems by Southleft, enabling AI agents to interact with design system data programmatically.
π‘ TJ addresses the common question of how to handle design tokens with MCP, demonstrating a workflow that bridges Figma variables with code-level token systems.
π‘ Cloudflare introduces Code Mode MCP, expanding the MCP ecosystem beyond design tools into infrastructure and deployment workflows.
π¬ TJ argues that most AI-meets-design conversation stays surface level, and digs deeper into how MCP and design systems can fundamentally change workflows rather than just add AI sprinkles.
π₯ TJ explores the Claude Code to Figma workflow, showing how bi-directional MCP connections create a new paradigm for design-to-code and code-to-design collaboration.
π‘ Demo video showcasing AI and design system integration workflows, including practical examples of using MCP servers with design tooling.
π¬ Brad's 2013 commentary on the WIRED 'Better Than Human' article β prescient thoughts on AI and jobs that resonate even more today.
π¬ Classic 2012 WIRED article about AI's eventual impact on jobs, shared as a reflection on how predictions from 13 years ago are now playing out in real-time.
π₯ Jesse Gardner shares work on applying design systems and AI to civic technology, showing real-world impact of these tools beyond corporate settings.
π₯ Design educator Christine Vallaure launches a course on creating with Lovable, demonstrating how designers can move further down the chain to create and ship real software.
π₯ Cursor launches agent skills feature. A community member built a custom skill that compares Figma and code to find token usage discrepancies β already found edge cases at scale.
π‘ TJ launches an extensive documentation site for the Figma Console MCP, answering community questions and providing inspiration for leveraging the tool within design system workflows.
π‘ Event registration for a Storybook team session on AI and design systems, featuring demos and discussion about the Storybook MCP.
π₯ Community member tested Anova plugin for comparing Figma components with React code. Found ~30% token savings with better output quality compared to passing raw Figma data. Reduces context saturation and boosts accuracy for design system audits.
π₯ TJ demonstrates write capabilities with design tokens as the anchor, showing what happens when context flows from design to code through token-first AI workflows.
π‘ Southleft's comprehensive 5-part article series on the Figma Console MCP, covering everything from basic setup to advanced AI-powered design system management workflows.
π₯ TJ demonstrates how Figma becomes an operable console β not just a canvas β with read and write access across projects, variables, components, docs, and pages from a single MCP-powered console.
π¬ Dan Donald explores integrating ethics into AI workflows, sharing a framework (available as a Gist) for ethical consideration when using AI in design and development. Built as a Claude Code plugin.
π₯ Anthropic launches Claude Cowork as a research preview β a desktop tool for non-developers to automate file and task management. Shared with excitement by the community.
π‘ Buoy is a design tool that caught the community's attention for its approach to AI-assisted design workflows.
π‘ Open source tool that generates UI compositions using your custom design system within Storybook. Works with React, Angular, Vue, Svelte, and web components. Demonstrated installing into Eddie Design System's web component library.
π‘ TJ demos the Southleft MCP for WordPress management β creating a full services page without ever logging into the WordPress backend. One console to rule them all.
π‘ A2UI (Agent-to-User Interface) explains what agent-driven UI is and isn't. Includes a LinkedIn post, article, and demo video breaking down how natural language + MCP tooling collapses the gap between intention and execution.
π‘ TJ breaks down A2UI (Agent-to-User Interface) for the curious β a short LinkedIn post + demo video explaining how agent-driven UI works and what it means for traditional GUI patterns.
π€ Brad references the Arrested Development model home scene as a metaphor for AI-generated code that looks great but might lack solid architecture underneath. A humorous but pointed reminder about verifying AI output.
π‘ A design style explorer for AI prompts β useful for experimenting with generating UI or visual assets with specific aesthetics. Relevant for AI-assisted design workflows.
π₯ TJ explores where UI is headed with AI workflows β how natural language + MCP tooling is collapsing the gap between intention and execution, making traditional GUIs optional.
π₯ Brad's thoughts following a session with the Storybook team, highlighting how AI and design systems unlock more collaborative workflows. Includes video clips of key moments.
π‘ Dan Donald's MVP of a new approach to design systems tooling β exploring AI-powered workflows for design system management.
π‘ Registration for the Storybook MCP webinar featuring Brad Frost and the Storybook team discussing how AI and design systems can ship production-grade code.
π€ Chris Coyier argues AI has a 'Jeopardy Phenomenon' β when AI generates code outside your expertise, you think it's fine; but when it touches your domain, you spot all the mistakes. A cautionary note about trusting AI output.
π‘ Molly demonstrates using Figma Make to automatically write design system documentation, sparking community discussion about AI-generated documentation for design systems.
π‘ Dan Donald explains agents, orchestration, and personas in the context of Claude Code β aimed at encouraging folks who haven't taken the plunge yet to explore AI development tools.
π₯ Storybook team posts about AI-generated UI using design systems, demonstrating how the Storybook MCP enables AI to generate components that respect design system constraints.
π‘ Dan Donald's guide aimed at getting folks who 'just prompt' to take the next step with LLMs β focusing on context management and more effective AI collaboration.
π₯ Official announcement of the AI & Design Systems course by Brad Frost, Ian Frost, and TJ Pitre.