Custom brand tools feel like the most compelling slice of AI right now
Working on a technical logo for a client, got tired of endlessly resizing a bunch of lil' rectangles. So I built a tool to play with variations
and animate it, of course
Yess love this! Taste is so inherently human.
Have spent a year exploring deeply the limits of true creativity within agentic systems. And like @charlicohen mentions, the key flaw with Ai is that what creates taste is so antithetical to how Ai is trained.
In this new world there is a bigger need than ever for spaces for human taste and creative POV’s to thrive without the noise
@benblumenrose This is great @benblumenrose! Was there any more detail to the design stack that wasn't included here? Seems like a lot of the new toolset for creatives such as Flora, Paper, Magicpath, etc didn't show up here?
Our 2026 Design in AI Report is now live!
This report is the culmination of thousands of people hours and many late nights to create what we believe is the most comprehensive, well-researched report capturing and synthesizing the state of Design + AI today.
While we used AI in many areas, a report like this still required deep thinking, grit, and humans coming together to do what they do best.
The final report spans nearly 20k words covering the survey results of over 900 people paired with dozens of qualitative interviews.
Over the coming months we will also release 7 beautiful case studies showing how top design teams are working on the ground featuring designers at @AnthropicAI, @framer, @linear, @NotionHQ, @Shopify, @SierraPlatform, and @stripe.
This work is a true labor of love to help guide a design community we hold so dear.
Link in the comments and please let us know what you think. Your feedback helps us shape how we will evolve this work over the coming years...
An interesting shift is occurring for in-house brand design teams:
Our job is becoming less about making every brand expression ourselves, and more about building the tools that let the company express the brand beautifully at scale.
The future brand designer is part designer, part product maker. It’s going to be fun to see what skills this unlocks, and the second-order effects it will have on the aesthetics of brands in an agentic era.
My team is already blowing my mind with the problems they’re solving alongside their best friend, Claude.
"Aim to be paid for your mind, not your hands."
Some of the sharpest thinking I've been close to this year. The /human + /agent split is the cleanest articulation of where brand work is heading
@littleplainsxo team is absolutely owning the way to embrace this change how we work creatively
Have a lot of people hitting me up for Ai talent - Ai Brand Designers, Ai Photo & Video Creators, Ai Designers. These are big brands looking to bring on exceptional creatives who really know what they are doing in the creative space.
Everyone I know is full booked - If you know of, or are a talented ai creative, hit me up!
Yung Lean steps into the world of GENER8ION—the project between French producer Surkin and director Romain Gavras—for an explosive double single and director's cut film.
"STORM I" & "STORM II" unfold across two chapters within the same all-boys school setting: the first volatile and disruptive, the second reframing that same environment through choreography by Damien Jalet, with "space intentionally carved into the infectious production to amplify the movement, drama, and physical expression,” according to a press release.
The director's cut is out today, and an alternative performance video for "STORM II" follows April 29.
Happy to share what we've found so far as it's pretty relevant!
- Basically, standard CLIP can't distinguish between semantics and aesthetics. It literally can't see the difference that your critics can articulate.
- DINOv2 (self-supervised, no language training) does a good job at separating visual style from content.
- A fine-tuned LoRA and a "Charm" layer are really successful at pulling apart aesthetic qualities that off-the-shelf models collapse together.
- The training signal is just grouping "these images belong together" expressed as https://t.co/8XsmdTrCkD channels. No annotation, no rating scales. Just curatorial instinct. Which is honestly the coolest part!
One result that might interest you:
Same person, same outfit, different lighting only... the trained model sees those as meaningfully different where CLIP sees them as basically identical.
It learned that editorial lighting is an aesthetically meaningful axis. Which is the kind of thing your critics reason about through language, but detected perceptually in milliseconds.
The interesting connection to your work is that our embeddings can measure that images are aesthetically different, fast and cheap. Your critic cards can articulate why. Feels like the two approaches meet naturally which is cool. structured theoretical frameworks as an interpretive layer on top of the perceptual separation.
Would love to compare notes sometime!