Discovery Is Not Marketing

A decade with Peter Gabriel's orbit, LEGO figures in 2014, ChatGPT in 2023, a Stable Diffusion competition, and what the whole arc reveals about visibility, signal, and what AI actually changes.

In September 2014, a LEGO figure I had built and posted on Instagram was spotted by Peter Gabriel's social media team. They emailed me. Within a month, Musical Brick existed.

In early 2023, I started using ChatGPT and messaged Real World's marketing team to share what I was seeing. In April 2023, Peter Gabriel partnered with Stable Diffusion to run an AI video competition. I entered. I learned seeds, algorithms, prompts, negative weights, sampling methods. I submitted something I was genuinely proud of.

It wasn't chosen.

These two episodes, nine years apart, both orbiting the same artist, taught me more about how discovery actually works than anything I have read about marketing, SEO, or content strategy. And together they form what I think is the most honest case study I have for what AI changes about visibility, and what it does not change at all.

TIMELINE

Act one: how discovery worked before AI

The fifteen months between June 2013 and September 2014 looked, from the outside, like nothing. A language blog. Regular posts. A small, niche audience. No viral moments, no strategy, no growth hacking.

What that period was actually doing was building signal. Not content, signal. The distinction matters. Content is what you produce. Signal is what accumulates over time to make your content credible, contextualised, and findable by the right people. Content without signal is a message in a bottle. Signal is what gives the bottle a current to travel on.

When the LEGO figure appeared in September 2014, it did not land in a vacuum. It landed on top of fifteen months of consistent, specific presence. Peter Gabriel's team monitored hashtags. They were paying attention to a particular corner of the internet, and the figure was specific and credible enough to be worth acting on. No algorithm gamed. No promotion run. Just the right thing, in the right context, seen by the right person.

The lesson I drew from that, slowly over many years, is that discovery is not something you trigger. It is something you enable, through accumulated signal, specificity of output, and the patience to keep building something real in a defined space long before anyone is watching.

Nine years later

Act two: what changed when AI arrived

In early 2023, I started using ChatGPT seriously. I found it genuinely remarkable, not as a replacement for anything I was doing, but as a tool, a way of compressing certain kinds of work. And because I had a direct line to Real World's marketing team from years of correspondence around Musical Brick, I did something that felt natural: I messaged them to share what I was seeing.

Not a pitch. Not a warning. Just: here is this thing, it is moving fast, it is worth paying attention to.

A few months later, in April 2023, Peter Gabriel announced a partnership with Stability AI, a competition inviting fans to create AI-generated videos using Stable Diffusion. The brief was open. The tools were new. The barrier to entry was, theoretically, zero.

I entered.

What followed was several weeks of genuinely absorbing technical work. Stable Diffusion is not a point-and-click tool, or it was not then. You work with seeds, which determine the starting noise pattern and make outputs reproducible. You choose samplers, which affect how the model navigates from noise to image. You write and rewrite prompts, learning which words carry weight and which are effectively ignored. You work with negative prompts to exclude what you do not want. You iterate through hundreds of outputs to find the handful worth keeping.

I found it fascinating in the same way I find LEGO fascinating: a constraint-based creative medium where the interesting work happens in the gap between what the tool wants to produce and what you are actually trying to make.

I submitted something I was genuinely proud of.

What the competition revealed

The entries that won were technically accomplished and visually striking. They were also, in most cases, made by people who had developed a more refined visual vocabulary within those tools, and had a clearer sense of what they were trying to say with them.

Which is to say: the competition did not reward access to tools. It rewarded accumulated craft within those tools. The barrier to entry was low. The barrier to distinction was exactly as high as it has always been.

AI lowers the cost of participation. It does not lower the cost of being worth noticing. Those are different problems, and conflating them is the central mistake most people make about what these tools change.

This is where the two acts of this story connect. In 2014, the LEGO figure got noticed because it sat on top of fifteen months of accumulated signal. In 2023, the competition entries that won were not just technically skilled. They were the output of people who had been building signal within AI and other creative tools for months or years before the competition existed. The mechanism was identical. Only the medium had changed.

The discovery gap, widened

Here is what AI has actually changed: the supply side of content has increased dramatically, while the signal side has remained just as slow and accumulative as it always was. More people can now produce technically competent creative work. The competition for the attention of the people who matter has intensified without the underlying mechanism of discovery changing at all.

Discovery still rewards specificity. It still rewards consistency over time. It still rewards the work that sits at unusual intersections and builds credibility in a defined space before trying to reach beyond it. AI does not shortcut any of that. What it does is make it easier to produce the appearance of those things without the substance, which makes the substance harder to find, not easier.

This is the discovery gap: the widening distance between the volume of content being produced and the capacity of any discovery system, human or algorithmic, to reliably surface what is genuinely worth finding.

The most durable signal is not the thing you make. It is the decade of specific, consistent work that gives the thing you make somewhere credible to stand when it arrives.

What I am building toward

I still use AI tools. I find them genuinely useful, for thinking, for research, for compressing certain kinds of work that used to take longer. I am interested in what they make possible at the edges: the unusual intersections, the niche applications, the places where a small amount of AI capability combined with deep domain knowledge produces something that neither could produce alone.

That is the territory I am most interested in watching. Not AI as a content factory, but AI as a signal amplifier, for people who already have something specific and real to say, and need better tools to make it findable.

The ecosystem I work across reflects that:

The full origin story, the language blog, the fifteen months, the Real World email, is at AngMohDan. The craft detail behind the LEGO builds is at Musical Brick. This is where I think through what it all means as the landscape keeps shifting.

I did not land a position in the Stable Diffusion competition. But I learned the tools, understood the medium better, and came away with a clearer sense of what AI changes and what it does not. That felt like the right outcome for someone who has been building signal slowly, in a specific space, for over a decade.

Discovery still works the same way. It just has more noise to travel through now.