Despite the challenges we faced as a company, I still believe this work represents an inspiring example of early creative AI exploration — back when the focus was genuinely on augmenting human creativity rather than replacing it. These projects emerged from a time before the current generative AI landscape, when we were more concerned with empowering artists and designers than automating them away.
The original text remains unchanged from
With love, Roelof
While we successfully build some interesting technology, and conducted a lot of user testing and product experimentation, if we're very honest with ourselves we have to admit that we never really managed to turn our efforts into a single engaging and enjoyable product. Instead we found ourselves helping agencies, designers, and brands as a service consultancy, effectively transitioning partly into an agency for creative work "dog-fooding" our internal tools, but without a real urgency about putting these tools out for a wider audience: a clear lack of product focus.
This, combined with disagreements about strategy and business focus with (now former) two other cofounders, made it a difficult journey for everyone involved.
In short: it's time to call it a day, and move on. We — I speak now for me, Sam and Tero — still very strongly belief that the new growing field of Creative AI is an exciting opportunity for developing human-centric technologies, products, services, and tools that aid creatives everywhere. And for myself I can say, perhaps a bit more humility and focus as well :)
This all doesn't mean that it has been a total waste. A lot of people working at creative.ai have told me in person that is has been the most intense, but also the most transforming experience of their career: many have found passions to pursue or career switches to make (designers turning developer, data scientists turning artists, etc), by and large made possible through the supporting culture we build all of us working together.
Next to this we have dreamed up, and developed many interesting "new" (new as in unlike what's around today, but not as much new for those willing to look back a couple of decades, to the early computing pioneers for example) ways of building user interfaces, interfacing with tools, or ui/ux principles and designs.
Inspired by work from Christopher Alexander idea of a "pattern language" and his series "The Nature of Order", we imagined a design tool that would operate on the level of "patterns", and cognition. The frame of thinking followed here is that a large portion of creative work can be seen as "mechanical execution", where the creative spark comes from ideas, intuition, insights, and experimentation. Could not creative work be greatly accelerated and become more "creative", if the execution was done more automatic or enhanced by AI? In this way, a system could steer in the right direction, putting you in the driver seat of making the choices and deciding on the parameters to explore, a shift towards curating/picking the results one likes, to then move forward on, and the cycle repeats, in a cybernetic fashion between you, the system and its feedback loop.
left: Alexander's 15 design patterns // right: our imagine design patterns, where structure is more formed by the first 10, and visual/stylistic choices by the 5 "styles"
One area where systemic design takes place is during the development (or "update") of a brand personality. Often such branding exercises revolve around around what archetype you are as a company/brand/person: are you more conservative or extravagant? more emotional or rational? more playful or serious, etc.
Join Medium for free to get updates from this writer.
Subscribe
Subscribe
For the above mentioned exhibit we imagined a generative branding system that lets visitors play around with similar binary personality choices. But instead of getting to a single main archetype we used the choices to position a user of the system into a specific spot on this "branding world map" where the outer extreme edges were formed by the more stereotypical archetypes (Hero, Jester, Explorer, Maverick, etc).
The user made its choices in the form of a DJ mixer, where the knobs (like rotary controllers on a midi controller), made you choose between particular attributes (eg Playful or Serious, Friend or Authority, etc) which together decided on you place on this branding landscape (or high dimensional latent space in neural net speak), which were then fed in real-time to a generative design system that generates a poster matching a style and composition. A larger scratch pad let you create variations around this aggregated theme, so a user could get a sense and feel about the particular “brand” and the possible variations contained in that particular place.
Four different "extreme" edges of a design space "in making", that are tailored to represent particular "archetypes", with some design criticism on how to improve the design quality further.
A quick and simple hack that lets you explore visual themes and the path in between these themes was born out of a Friday hack, we referred to as the "weirdness-engine". The idea is that
However, we quickly saw that the value of blōma extended beyond this initial step of generating ad variations. Blōma started looking more like a way to accelerate visual design work in general, leaning on a few key ideas:
Systemic design. Creating variations based not merely on static templates and placeholder, but systemic rules that give rise to large variation spaces that adapt to different formats and different types of content.
Generative design. Instead of starting design work from a blank canvas, starting from large design spaces already generated for you, where your design work then becomes a matter of navigating and filtering the design space, rather than building things from scratch.
Embedded brand style guidelines. When you have an existing brand, you usually have visual guidelines for expressing the brand in terms of typography, colours, and image treatments. With blōma, you can embed these guidelines into a living design tool, allowing you to quickly produce new content that is on-brand and high quality.
Finally, I'm very thankful for everyone who was part of our journey, especially Sam and Tero who were part of creative.ai until the bitter (sweet?) end.
And from the three of us: thanks to Petra, Ruby, Nikola, Damir, Denis, Dan, Stéphane, Carmen, Gloria, Ivan, Dina, Sorin, Miguel, Mikkel, Andrei, Imanol and Sofia for being with us for the months or year+.
And big thanks to our investors, Max of Sunstone (now Heartcore), Suzanne of Local Globe, and Jan & Stefan of Discovery Ventures. Thanks for all the support and help these past years!
Last, but not least, I want to thank Alex and Samim for starting this journey together, and for all the time we've shared, and lessons learnt from each other.
Business, governance, and adoption-focused material. Real-world implementations, case studies, and industry impact.
