It's Monday. Here's what changed in AI this week, and what it means for your design work:

Back in Cyprus. Ready to lock in for the next month and a half.

I spent the past few weeks mapping out every task I do on a daily basis - and the plan is simple: delegate all of it to Claude Code. I also picked up a Mac Mini that I'm setting up as an always-on automation server. The goal is to get completely out of operations so I can focus on what actually moves the needle - growing the business and creating content.

On the product side, both Limora and Stage are making real progress. We're still heads down on development, not pushing marketing yet - but the results so far are exciting. I can see the potential, and that's enough to keep me locked in.

This is the biggest Figma update since Auto Layout.

Figma just launched open beta for AI agents that can design directly on your canvas. Through the new MCP server and a "Skills" system, any compatible agent - Claude Code, Cursor, Copilot, Codex, and more - can now create and modify frames, components, variables, and auto layouts. Not through screenshots or plugins. Directly in your Figma file, with full awareness of your design system.

The way it works: you install Figma's MCP server, connect your preferred AI coding tool, and the agent gets a use_figma tool. From there, it can generate designs, modify existing ones, and even screenshot its output to self-correct. Skills are simple markdown files that teach agents how to behave on your canvas - your conventions, your spacing rules, your component usage patterns. Write them once, agents follow them across every project.

Nine community-built Skills shipped on day one. Uber built one for generating screen reader accessibility specs. Edenspiekermann made one that connects designs to system components. Warp built a workflow orchestrator.

Starting this week, AI agents can design directly on the Figma canvas. Not through a plugin. Not through a screenshot. Directly on your file, using your components, your variables, your spacing tokens.

Here's the thing - this isn't about AI replacing designers. It's about speed. The designer who can describe what they want to an agent and refine the output in 20 minutes will outpace the one who builds every frame from scratch in 3 hours. Same quality. A fraction of the time.

The skill gap in design just shifted. It's no longer about how fast you can push pixels. It's about how well you can direct an AI to push them for you. Design systems just became machine-readable in a way that actually changes workflows. An agent can now read your codebase, understand your design system, and build screens using the right components with the correct spacing. The handoff gap between design and development just got a lot smaller.

It's free during beta. Set up the Figma MCP server, connect Claude Code or Cursor, and try generating a screen from your existing component library. Start with something simple - a settings page or a dashboard card using your design tokens. You'll know within 10 minutes whether this changes your workflow.

If you've wanted to add 3D to your web projects but couldn't justify the time investment, Omma just removed the excuse.

Omma is Spline's new AI creative studio that builds interactive 3D websites, apps, and WebGPU experiences from text prompts. Describe what you want, and multiple AI agents work in parallel - one handles code generation, another creates 3D meshes, another generates images. The result is a fully deployable project, not just a prototype.

It supports data pipelines too. Feed it CSV, JSON, or 3D files (GLB, OBJ, GLTF) and it builds live, data-driven experiences. Deploy to production with a custom domain directly from the platform.

If you've ever spent hours going back and forth between design tools and code, this workflow cuts that down to minutes.

Here's how to go from concept to deployable prototype in three steps.

Open Figma Make and describe the app you want to build. Be specific about the purpose and style.

Then add a visual reference to guide the direction. This is what separates generic AI output from something that actually looks premium. Run the same prompt with different inspiration images to get multiple concepts you can compare.

Pro Tip: The quality of your reference image determines 80% of the output quality. Spend an extra minute finding the right one.

Copy and paste the best concept into Figma. This is where you do what AI can't - take time to refine the typography, clean up the layout, and balance the colors.

This is the step where you remove the AI slop. Adjust the spacing, swap generic fonts for intentional ones, fix the hierarchy. The AI gets you 70% there. The last 30% is what makes it look like a designer actually touched it.

Paste the refined design back into Figma Make and ask it to update the screen based on your refinements.

The best part? Figma Make generates a working prototype. Not just a mockup - real code. You can export it or publish it directly as a live site.

That's it. Concept to deployed site in three steps. The AI handles the heavy lifting, you handle the design decisions that actually matter.

Before you go we’d love to know what you thought of today's newsletter to help us improve The Logiaweb experience for you.

💌 Know a designer who should be using AI smarter?
Forward them this email. Or just send them to logiaweb.net to join.

See you next Monday,

Say hi on X, Instagram or YouTube