A labor of love.

A bit about me. I've been an avid fan of new technology, and a firm believer in testing new tech hands-on whenever possible. When Cursor came out in the spring of 2023, one of the first things I did was build a vector-based DAM. My reasoning was that converting images to vectors and storing them in a database would yield superior findability and unlock capabilities that weren't possible before.

I tinkered with it, played with vector cosine-similarity clusters, and did some interesting things. But it was a toy — a prototype, just to get some ideas into a sandbox and explore what was possible. I developed it a bit further as a local proof of concept, but agentic coding at the time meant that building anything serious still required nearly as much engineering discipline as building the real thing.

In the spring of 2026, I revisited agentic coding and was genuinely impressed at how far it had come. One of the ideas I decided to entertain again was the vector-based DAM. And this time I tried to push the limit — incorporating a ton of ideas and concepts I'd seen mentioned or requested by other DAM users, just to see what would be possible. In my day job I work with enterprise-class customers, so I wanted to make sure that whatever I built was clearly set in a class that doesn't compete in the enterprise space.

A bit of a frankenstein.

The result is a bit of a frankenstein: enterprise AI features in a DAM built for freelancers and small agency teams. It has been an incredible learning experience — building a SaaS application from scratch, drawing on my cloud-architecture background, my web-development heritage, and my enterprise-DAM experience. VDAM is a labor of love. Building it has been an obsession, and it has induced a bit of temporary AI psychosis: I've found that the mental load is now more about making decisions and solving problems than the minutiae of low-level code debugging. A lot of the features and capabilities I incorporated would normally require detailed Jira stories, epics, and research spikes. But the implementation time is so fast that it has been a seismic shift in what I formerly knew as the software development life cycle.

What a DAM should be.

I hope you enjoy VDAM. The first release is built on a lot of my personal convictions about what a DAM should be. It should make the content the star of the show, so I made the UI as minimal as possible — to give as much attention and screen real estate to your content as I could. It should allow easy, intuitive operation, so the prompt bar lets you do much of what you need a DAM to do. And sharing should be as customizable as possible: I know you're proud of your content, you want to share it, and the vehicle for sharing it should let you express your own brand — your voice, your look and feel. I hope I've provided enough tooling to enable that.

It should be innovative.

One of my proudest features is the auto-growing tag trees. Today, DAMs are starting to offer custom-prompted metadata. I take that idea three steps further: I tag your content based on your custom metadata prompt using an existing tag taxonomy — and not only that, if there isn't an entry in the tag tree, VDAM inserts it into the tree as it ingests your content. No one else has this. So I patented it. For the DAM nerds out there (like me), it allows hierarchical auto-tagging in a controlled vocabulary that grows and maintains itself. If that last sentence got you excited, you're my kind of people, and we have to go out for beer or wine or drinks sometime.

I intend to write more, but I just wanted to get this first one out. Feel free to write to [email protected] — or, in the app, prompt your 'feedback' and it'll get to me. Thanks!

Field notes from VDAM — a labor of love, written by a human.