Jaden Kwan

Founder of Tastefully and Tastefully Labs. Waterloo Engineering opt out. Best Ethereum Hack at JAMHacks, 2nd place at ETH Toronto. Professional actor since childhood, from Chucky to Finch and Midland at the Golden Horse Awards.

I bootstrapped Tastefully from zero to ~$300K in revenue in its first four months with no outside capital. We make launch films and run launches for AI companies backed by Y Combinator, SoftBank, and Radical Ventures, and our work has driven more than 25 million organic views. We led the launches for Dedalus Labs, Oro, and Yutori; Dedalus's seed film, Cleared for Takeoff, passed a million views. The Village Voice wrote about how we bring film direction to a startup launch, and Distractify about startups becoming their own broadcasters.

A year ago I had no network in San Francisco. I took a one way flight, slept on couches and in closets, ate at founder events, and started cold pitching launch films to startups. That became Tastefully.

I run Tastefully like a software company, not an agency. We built an internal agentic pipeline that handles the work around the creative. Agents do the repetitive work; editors spend their time on taste.

Building that pipeline is how I found the limits. Agents could automate everything around the edit, but not the edit itself. That gap is why I started Tastefully Labs.

Models can write code, but they still can't edit a good video.

Coding has tests. Creative work has two halves. One half is checkable: is the clip the right length, are the captions there, is the audio clean. The other half is taste: pacing, the hook, which shot to open on, whether people actually watch it. Today's agents struggle with both, and there's almost no good data to teach them.

I'm in an unusual position to fix that. Every week, real editors on our team do real professional work in Premiere for real clients. With Tastefully Labs, we're turning that work into training data: the raw footage, the brief, every edit decision along the way, the final cut, the client's feedback, and how the video actually performed once it was published.

From that we're building benchmarks that measure whether AI can do professional creative work, and environments where models can practice until they can. We're starting with video editing in Premiere, then After Effects, Photoshop, Figma, and beyond.

Most creative datasets stop at input and output. Ours captures the whole chain: footage, decisions, actions, edit, critique, revision, distribution, and real audience results. The studio creates the work; the lab turns it into intelligence.

My life follows one pattern: I get obsessed with something, go all the way in, and end up doing it for real.

In high school it was robotics and crypto. I was lead programmer on our VRC team, wrote a curriculum used to teach 50+ students, and taught myself smart contracts well enough to win hackathons.

Then it was film. I've acted professionally since I was young, and Finch and Midland, alongside Anthony Wong, took me from VIFF to the Hong Kong Asian Film Festival and the Golden Horse Awards.

After high school I spent a year landing in cities where I knew nobody: Vancouver, Mumbai, Taipei, Hong Kong, Shenzhen, Singapore, Johor Bahru. In Mumbai, the people I met became the post production team that powers Tastefully today.

AI will make building almost free, which makes taste the scarce thing. The next generation of models needs to learn taste from the people who have it. I want to be the one who teaches them.

I read constantly: business history, biographies, film, psychology, advertising.