FLORA

AI creation runs top down, so it needs a new interface. Traditional tools build media piece by piece, layer by layer. With AI you ask for a rough version and then chisel toward pixel perfection, so FLORA built an infinite canvas where text, image and video blocks chain into workflows.

Weber Wong points out that Adobe layers date to 1987, video timelines to 1991, and that Figma is a web version of the same model. As a banker he modeled a business in two days in Excel, but at NYU's ITP one creative output took weeks across 10 tools. He calls the canvas hard to design but says "there's really no other choice for the most powerful tool."

Subversive▶ 42:59

FLORA grew out of node tools Wong built for his own art. At ITP he made real-time AI paintings from a webcam feed. Others asked how, so he built them a node-based tool, which became the first FLORA. Node systems like TouchDesigner and Blender were already common at ITP, which made a connected canvas feel natural.

Rising usage was costing him money. He first considered an agency where one creative does the work of 100, but lacking agency experience he chose to sell the tool to every creative team. FLORA started about two years before the recording and launched about nine months before it.

Subversive

Collapse per-model nodes into a few blocks users actually think in. The alpha had a node for every model and was too complex and technical. FLORA rebuilt it from scratch around just text, image and video, so connecting text to image brings in all text-to-image models automatically.

"Creative professionals do not care about [the model]. They care about having a great video." The simpler abstraction also suits the planned control and agentic features, which are phases two and three on top of the current product.

Subversive▶ 12:26

Build the layer above the models instead of competing on outputs. Wong expects the obsession with single model outputs to fade as professionals ask how models fit their workflow. FLORA calls itself the creative computing layer, comparing it to Microsoft building BASIC, DOS, Windows and Office on top of PCs.

Next steps are making visual programs easy to build and share while adding control. Wong also pitches FLORA as "Harvey for creatives," a power tool for professionals.

Subversive▶ 42:59

Product/market fit came when users turned their process into workflows. After launch most people used FLORA to explore concepts fast. Fit arrived once users codified whole processes, such as full brand identity workflows; average nodes generated per user then skyrocketed and retention curves started to smile.

Early signals were qualitative: top agencies wanting into the alpha and Pentagram generating hundreds of logo variations at once. Lionsgate tested 1,000 story beats as videos before greenlighting a film, and a "generative brand script" can deliver a brand identity in two weeks instead of six months and hundreds of thousands of dollars.

Subversive▶ 19:02

Track a depth metric that separates power users from casual ones. FLORA's north star is nodes generated, not weekly active users. Users who adopt it as a creative system have 100 to 200x the usage of those treating it as a better Midjourney.

Weekly it also watches signups, conversion to paid and weekly churn, and still tracks WAU at lower weight. Wong cites hundreds of thousands of light users and roughly 1,000 deep users, and estimates about 10,000 people or teams worldwide work this way; none of these figures, or revenue and retention numbers, were verified.

Subversive▶ 37:28

Ask what your users will need in 5 to 10 years. Wong never asked what others weren't doing; he asked what creatives will literally need in 5 to 10 years. Almost no one believed him at first, which he reads as the mark of a non-consensus view that turned out right.

He compares FLORA to Nike, founded by a runner for runners, and says no other tool in the space is built by creatives. He sees the category as year one of an expected 10 to 15.

Subversive▶ 21:44

Launch with a sharp premise about an underserved group. FLORA launched saying current AI creative tools are "built by non creatives for other non-creatives," and it would build one for professionals. The launch went viral among professionals right away.

At launch there was no overlap between AI-tool users and professionals; aiming at professionals instead of hobbyists is what made it take off.

Subversive▶ 12:57

Let feedback adjust the margins, not the core vision. FLORA has never pivoted. It adds quality-of-life features only when it hears the same need several times in one week, and core innovations come from first-principles thinking.

The first version in February had the right skeleton, needing only small additions. "The core innovations do not come from listening to all the noise and then trying to turn that into a bunch of small features."

Subversive▶ 35:55

Hire customers' AI-savvy creatives to onboard other top teams. FLORA hires creatives from customers like Pentagram to work inside other creative teams and show how AI fits their process, modeled on Harvey's legal engineers.

A new power tool needs change management the founder can't deliver at scale, and Wong says these hires "do a better job than me" at contextualizing it.

Subversive▶ 30:53

Project-level sharing turned usage into organic signups. After FLORA added project-level sharing, more signups came from organic sources rather than top-of-funnel marketing.

No figure was given. Wong plans more growth loops and sharing to spread power users' knowledge within teams.

Subversive▶ 39:47

Own the education for a new way of working. FLORA hosts generative media talks with Pentagram and A24, published a Field Guide as an operating manual for the future creative team, and runs a creative partner program where creators share workflows.

Word of mouth is the biggest growth driver, with steady weekly signup growth and no sophisticated marketing.

Subversive▶ 30:20

Narrow to verticals where adoption and output quality are highest. FLORA started open to every creative field but focused over time on fashion, CPG, and brand and marketing, where adoption was strongest and AI reached production grade more often.

Brand and marketing is the biggest use, fashion, CPG and shoe design second, film and VFX third. Nike uses it to visualize shoes as well as for marketing; users range from Netflix and Lionsgate to a Montana tire shop owner making his own ads.

Subversive▶ 27:55

Teach the core skill in onboarding, then motivate with a big result. FLORA swapped an onboarding of big example workflows for one teaching how to connect text to image to video, which has gone well. The plan is to show an impressive result first, then teach the skill.

A head of product hired a couple of months earlier, who had scaled Design Lab to 50 million in revenue, made onboarding noticeably better and added the metrics expertise the founder lacked.

Subversive▶ 40:00

Explain a modular tool in each team's specific terms. Because FLORA can do many things it is hard to explain, so it learned to describe its value specifically for each kind of creative team, likening the problem to Notion's early days.

"Poetry is about expressing yourself, whereas copywriting is about expressing to someone else."

Subversive▶ 34:55

Ship early and try cheap tweaks, even pre-scale. FLORA got stuck adding scope and polishing, and improved by shipping, testing and iterating. It also switched some buttons from white to green, which Wong says has been helping conversion.

No number was given for the button change. FLORA runs few A/B tests so far but plans more, especially on growth loops. "Product is like an artwork that is continually built upon for decades."

Subversive▶ 53:02