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

There was a moment this week where I looked at a set of screens Stage had designed and thought, that is it, that is the thing we have been trying to build for months. Not because the wireframes were better than last month. Because for the first time everything behind them was in the right place.

The problem was always packaging. Stage does research, strategy and process before it designs anything, and we never had a clean way to carry all of that into the wireframes. We tried a lot of things. What finally worked was exporting the full body of work into a dedicated workspace and designing from there. Paired with GPT-6 or Fable 5.1,the results have been incredible.

The setting changed too. I moved from Larnaca to Paphos this week, the west coast, where the nature is on another level. Saturday was a quad tour through the hills and a few swims. The rest of the week was the same routine as always, just with a better view.

The new Stage landing page is up and it explains the product properly. I am also doing demo calls, so if you are curious, book a call here.

For the first time, I do not see a reason to open a different image model for asset generations.

OpenAI shipped ChatGPT Images 2.5 on Monday, and the two features that matter are about how you direct the model, not how pretty the output is. Sketch lets you type @Sketch, draw a rough layout inside ChatGPT with a mouse, stylus or finger, add a short brief, and the model uses your drawing as the composition reference. Comments let you drop a note on one area of the image and change only that, and earlier edits now survive later ones.

It is also up to 50% faster, holds the likeness of people, products and pets across edits, and adds Templates for flyers, posters and product photos. In the API there are two models: GPT-Image-2.5 Flare, the fast default, and GPT-Image-2.5 Sunburst, the slower premium one built for tight control across repeated edits. Both cost exactly twice GPT-Image-2.

Here's the thing. Flare and Sunburst are not two options, they are one workflow. Flare is cheap and fast, so you use it to iterate: ten layouts, five color directions, three crops, without thinking about the bill. Sunburst is slow and precise, so you use it once, on the direction that won, and it holds the details through every revision. That pairing is why it almost makes no sense to reach for other general models anymore.

We plugged both into Limora AI the day they shipped, and the experience is a clear step up. Faster first drafts, and finals that do not drift when you ask for one more change.

That said, some models still own their lane. Krea is still where I go for creative exploration when I want to be surprised. Recraft 4.1 is still the pick when consistency across a whole set of assets matters more than anything else. Images 2.5 does not replace those. It replaces everything in between.

Every launch needs a video, and every video takes a day you do not have. You have the screenshots, the clips and the copy, and somehow it still ends up as a weekend in a timeline you barely know.

EndFrame is a Mac app where Claude, ChatGPT or Grok edit the video for you on a real timeline. You describe the launch video, demo or social clip, drop in your images, clips and audio, and the agent writes the scenes, renders the frames, then checks its own work against a 33-rule quality gate for things like margins and text contrast before you ever see it. You can scrub, trim, go back to any version and export in 1080p or 4K. It runs on your existing Claude or ChatGPT subscription, no API keys, and it is free during early access. macOS 15 and Apple silicon required.

I tap my desk and Claude Code opens. Holo is a free, open-source Mac app that listens to your desk through the built-in mic and maps taps to actions. Here's the five minute setup.

💌 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