This website uses cookies

Read our Privacy policy and Terms of use for more information.

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

  • 🧪 What I'm Building: Back from Ibiza, and a new Cursor partnership

  • 🚨 Big News: Figma Make gets a Properties Panel and Annotations

  • 🤖 Prompts Inspiration: 3 prompts for editorial art, brand visuals, and product shots

  • ⚙️ Tool of the Week: Mirafold

  • 🛠️ Tutorial of the Week: Raw footage to a punchy Reel in 5 minutes

  • 🔗 Quick Links: 5 things that caught my eye

WHAT I’M BUILDING

Behind the Scenes

Back in the studio this week after switching off in Ibiza. And honestly, I needed the reset - but I also let content slide while I was gone, so job one this week was getting the machine running again. Batch filming, scripting, back into a real rhythm.

The bigger news: I signed a partnership with Cursor. I've been heads-down building the content for it all week. It's tied to a new product release, so I can't say much yet, but it's the kind of thing I'd have wanted to see when I started learning how to use AI.

On the Klime side, we started rebuilding the whole website. We shifted our positioning recently, and the current site doesn't reflect where the studio actually is anymore.

BIG NEWS

Figma Make Just Killed the Prompt-Only Workflow

The prompt-only era of AI design is ending, and Figma is the one ending it.

Last week Figma Make got two things it was badly missing: a real Properties Panel and Annotations. The Properties Panel brings the Figma Design controls you already know - spacing, typography, line height, letter spacing, opacity, z-index, border style - straight into Make, plus a DOM tree view and batch-editing across multiple instances. Annotations let you mark an area on the canvas and describe an interaction or animation for the agent to build, like a full-screen nav overlay when the menu button is clicked.

The smart part: your edits stage in the chat panel credit-free until you hit apply. So you can line up a whole batch of changes, review them, then commit once.

My take

For a year the pitch for AI design tools was "just describe what you want." It sounded great and it was exhausting. Describing a 4px spacing change in a sentence is slower than dragging a handle, and you're never sure the model heard you right. This update quietly admits that. Direct manipulation is back, and the AI handles the part it's actually good at - writing the code underneath. That's the model I think wins: you click, it codes. It also uses far fewer tokens than prompting the same change, which matters now that everyone is watching their credit meter.

How to use it

Start a fresh Make file (the Properties Panel only works on new Make designs for now). Build a rough layout from a prompt, then switch to Edit in the top-right and fix the details by hand instead of re-prompting. Save the annotations for the things you can't click - motion, hover states, logic.

PROMPTS INSPIRATION

3 Prompts for Editorial Art, Brand Visuals, and Product Shots

Painting-style digital illustration

Vibrant painting-style digital illustration of a modern city skyline at sunset. Thick brushstrokes and visible paint texture, palette-knife technique. Warm peachy-orange...

Timeless luxury advertising visual

Product Archive: hero a single iconic product as a giant hyper-real sculpture inside a dramatic branded world, cinematic lighting, bold tagline typography...

Premium food photography

Top-down premium food photography of a glass jar of chili tomato chutney on a dark stone slab, surrounded by fresh tomatoes, red chilies, black...

TOOL OF THE WEEK

Mirafold

If you use terminal coding agents but hate staring at a wall of text, this one's for you. Mirafold is a browser interface for Claude Code, Codex, and Gemini CLI with a generative UI layer on top. Instead of raw terminal output, your agent answers with live cards, charts, dashboards, task lists, and diffs you can actually read and pin.

It's open source, you bring your own key or run it fully local (Ollama, LM Studio, vLLM), and your keys never touch the browser. You can even pair your phone or tablet with a QR code and pick up the same session anywhere, end-to-end encrypted.

How I'd use it

  • Design-to-code reviews: watch an agent build a component and read the diff as a clean card instead of scrolling terminal noise.

  • On the go: pair your phone and check on a long-running build from the couch.

  • Local and private: point it at a local model for client work you can't send to the cloud.

TUTORIAL OF THE WEEK

Raw Footage to a Punchy Reel in 5 Minutes With AI Producer

Editing short-form is where most people lose hours. This week I ran raw footage through AI Producer by OpusClip and had a finished, captioned Reel in about five minutes. Here's the workflow.

Step 1: Upload your footage

Step 2: Describe the edit

  • Tell it what you want in plain language, like "cut this into a punchy 30-second Reel"

  • It works like a producer, not a template: animated captions, motion graphics, zooms, even B-roll pulled from the web, all shaped by your words

Step 3: Tweak and export

  • Want changes? Say it in the chat and it reworks just that section

  • When it's dialed in, download and post

QUICK LINKS - This Week's Radar

  • Figma AI admin controls - Admins can now set custom AI credit limits per user, so one person can't burn the whole team's credits.

  • Motionly - An AI-native motion graphics editor that generates animation projects you refine on a real timeline and canvas.

  • A11y Agent - Grammarly for accessibility: audits any page against 90+ WCAG rules and suggests copy-paste fixes in your browser.

  • Reignat Analytics - Ask questions about your website traffic in plain language and get instant answers.

  • Swift AI SDK - Add streaming chat, agents, structured output, and live voice to iOS and macOS apps.

💌 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,

Adrien Ninet

Say hi on X, Instagram or YouTube