Jeremy Jackson

For years, we have been told that good design is about making things easy to see. We obsessed over the perfect shade of “Retina Blue,” the springiness of a swipe, and the satisfying “pop” of a notification. But as we move deeper into 2026, the industry is waking up to a radical truth: the ultimate user interface isn’t a better screen, it is no screen at all.

3

Big news from Anthropic: Claude Design just launched. Imagine describing a product prototype, a pitch deck, or a one-pager in plain English, and having a polished visual ready in seconds. On April 17th, Anthropic made that real. What it does: • Generates polished designs from natural language descriptions • Reads your team's codebase and design files to automatically apply your brand system • Direct integration with Canva (export) and Claude Code (handoff to production) • Export as PDF, PPTX, HTML, or a shareable internal URL Market reaction said it all: Figma dropped ~7% on the day of the launch. The real shift here isn't "AI can design now", other tools already do that. It's that Claude Design closes the loop: from a rough idea, to a visual, to production-ready code, all within the same AI ecosystem. That end-to-end workflow is the game changer. Worth watching! 🔗 Official announcement: https://lnkd.in/dyrUZDrn

9

Anthropic just gave Claude the ability to design. The demos look incredible. Most of them won't matter to our business. Claude Design lets us describe what we need in plain language and get polished slides, wireframes, prototypes, and marketing collateral. We can export to popular formats or hand off directly to Claude Code for development. It's genuinely impressive. And that's exactly when I've learned to slow down. Here's what I think is actually useful for operations: → Internal decks that used to take hours now take minutes → Quick prototypes for stakeholder alignment before investing in real design work → Draft collateral giving my team a starting point, not a finished product → A tighter loop between concept and build when paired with Claude Code Here's what's demo noise: → The idea that this replaces our design team → Treating AI-generated visuals as production-ready → Adding yet another tool without asking: does this solve a bottleneck we actually have? That last question is the one I keep coming back to. AI tool announcements are arriving weekly now. Claude Design. Cloudflare's new Agent Memory service for persistent AI agents. Google embedding AI directly into Chrome. Specialized tools for everything from knowledge management to autonomous business agents. The pressure to "do something with AI" is very real. But the pattern I keep seeing is the same: teams adopt tools based on how impressive the demo looked, not based on where the actual operational friction lives. The question that cuts through the noise isn't "is this impressive?" It's "what bottleneck does this remove?" For Claude Design specifically: if our team spends real hours on internal visual communication, slides, wireframes, one-pagers, this is worth testing seriously. If that's not where our time goes, it's a cool demo. We bookmark it and move on. The most valuable AI skill right now isn't adopting every new tool. It's learning to evaluate them through an operational lens: what does this cost, what does it replace, how does it integrate, and who on my team actually needs it.

12

AI promised to level the playing field. Instead, it just raised the bar. If you can’t spot the gaps, these tools don't make you more capable. They help you fail faster at scale. The bar for "great" remains in the hands of people who know their craft.

22

3 Comments

I ran the same prompt through Google Stitch and Claude Design to see how two of the newer AI design tools actually compare on a real mobile app brief. Stitch came back faster with multiple screens and a full colour palette. It's free, which makes it an easy starting point for anyone testing the water. Claude Design took longer, but watching it work you can see why. It's coding underneath the design. Worth noting it's still in research preview and available on Pro, Max, Team, and Enterprise plans. Both are useful depending on what you're actually trying to ship. Which one are you reaching for?

33

21 Comments

Claude Design dropped was recently released every AI channel is showing the official examples and calling it a day we did something different we pointed it at Werkflows scraped our own website fed it into Claude Design and got a complete pitch deck presentation back then we took the output and codified it in Claude Code turning the design into actual code that runs design and code converging in one workflow what used to take a designer plus a developer plus a week of back-and-forth now happens in an afternoon if you are building anything that needs visual output - slide decks - landing pages - proposal templates - internal training materials the manual version of this work is dead the question is not whether AI can do the design work it can the question is whether you have the workflow to take design output and ship it most agencies are still doing this work by hand hiring designers, hiring developers, paying for tools that overlap, waiting weeks for revisions the whole stack just collapsed into a prompt and a deploy stay ready full walkthrough of the build on the channel https://lnkd.in/g-6akVRt

2

What if we could understand the secret sauce behind Claude Design’s Generative UI and use it ourselves? After Anthropic unveiled interactive widgets in Claude (sliders, charts, real apps inside chat), we didn’t just explore it, we’re rebuilding it. I have tried Reverse Engineering the design system behind Claude Design’s Generative UI: • how show_widget streams HTML into the interface • how read_me loads design context on demand • how everything renders progressively via direct DOM injection This repo is our attempt to recreate that entire experience in the open, with design tokens, widget patterns, and the foundations of generative UI. Learn More at https://lnkd.in/g6QEf3d9 If you’re building AI interfaces, this is a glimpse into where UX is heading. Looking for collaborators and feedback 👇 OSS at https://lnkd.in/gegMqu3k

12

I just launched a new blog focused on brand design workflows, AI and the systems required to make generative tools usable at scale. It’s about moving past prompts and aesthetics and into constraints, production logic, and deterministic design systems that actually work in real organisations. Brand design breaks when generative AI is introduced without constraints, In the first article I break down why I think generative AI is inherently non-deterministic and without rule-based systems, it ends up injecting entropy into brand work. The role of the designer is shifting from asset maker to system architect: encoding visual logic, enforcing constraints, and forcing AI to behave deterministically at scale. I would love to hear your thoughts about the concepts I discuss in here and how you have been applying Generative AI in your brand design workflows.

43

Tool design is the most important aspect of using agents. This underscores the need for Model Context Protocol - when you want agents to work for your business, you want custom tools, and MCP is how you make that happen.

7

Played around with Claude Design today. Research preview, so rough edges expected, but the core is genuinely impressive. Describe what you want, get a working design, refine through conversation. It reads your codebase and design files on onboarding, builds a design system from them, and keeps output on-brand by default. The ceiling for non-designers is going to move. Founders, PMs, and marketers who've always leaned on designers for early-stage concepts now have a faster path to something shareable and testable. Not good news for Figma, Canva, etc. in the pre-AI design stack. Still early. But directionally significant. https://lnkd.in/eQx9DHeC

7

4 Comments

Recently started experimenting using .md files with claude code, and then saw a job listing wanting a designer who could create/edit .md files. I wonder if design systems all start having companion .md files soon to use with AI-tools, or if .md files become the core of design systems?

The classic design process is evolving and fast. For African tech innovators, this is a wake-up call: the way we build products, design interfaces, and solve problems is changing globally. Leveraging AI, thinking beyond traditional processes, and embracing new design archetypes could be the key to creating world-class products from Africa. 🌍✨ Tech is no longer about following established playbooks, it’s about adapting, experimenting, and innovating faster than ever. #AfricanTech #Innovation #Design #AI #ProductBuilding #FutureOfWork

1

Some initial reactions to the release of Claude Design: 1. Anthropic is setting the standard for what pace of innovation is possible for a truly AI-pilled organization 2. Anthropic is using their AI capability not to reduce cost, but to significantly increase their ambition 3. No product building on top of LLMs is safe from direct competition from Anthropic itself: I’m looking at all the prototyping tools in the space 4. Even the historic incumbents like Figma should start wondering are they going AI native enough to compete in this new world https://lnkd.in/gjwWCQ-m

113

13 Comments

I've been playing with Claude Design since it dropped yesterday (Anthropic's 5th product launch in the past two weeks), here's the summary: This is a potentially game-changing tool but it is currently crippled by bugs and an absolutely brutal usage limit. It has real potential to finally fix the nightmare of aligning elements in AI-generated code, which is a massive time-sink for developers. The ability to automatically create a design system from your existing codebase is a killer app (and a huge threat to tools like Figma), if it works as advertised. Plus, it has its own separate usage meter, so it won't drain your regular chat and code limits. However, the "Research Preview" label is doing some heavy lifting. The biggest complaint by far is the insane usage burn. Users on Pro and even Max plans report getting locked out for up to a week after less than an hour of tinkering. It's also buggy, with users experiencing errors, failed exports, and the classic LLM problem of not truly "seeing" the layout it's creating. You can try to access it at claude.ai/design, but it's a staggered rollout. It's web-only for now and appears to be available to both Pro and Max users who have been granted access. Have you tried the tool?

5

5 Comments