Voice Is the Last Moat

I joined Dr. Niklas: Venture Grade to talk about something I have become increasingly convinced of while building Bloomberry: voice is the last moat.

It is a short phrase. But there is a real argument behind it, and it is worth unpacking — because I think a lot of founders are about to make a very expensive mistake with AI.

If you want to watch and read the full episode, the complete episode notes and transcript are here.

AI Made Content Abundant. It Also Made It Forgettable.

Here is what I keep seeing. A talented senior person — VP at a company, founder, executive — decides to start posting on LinkedIn. They turn to ChatGPT, give it a brief, and it outputs something that starts with something like: "Most teams think the problem is X. But really it's Y."

And I know immediately. Not because I can prove it was AI. But because I have read that exact sentence structure a hundred times this week. The model has a default. It runs the same emotional arc, the same contrast setup, the same motivational kicker at the end. And I scroll past.

That person is talented. They have real insight. But by outsourcing the surface of their thinking to a generic model, they have made themselves sound like everyone else.

This is the problem Bloomberry is trying to solve. Not "write more content faster." But — write in a way that actually sounds like you, at scale.

What I Mean by AI Sentence DNA

Bloomberry's research team has catalogued over 7,000 cadences, phrases, and structural patterns that AI models consistently use. We call this AI Sentence DNA.

It is not just about words like "delve" or "leverage." It is about repeated emotional logic. The way models build arguments. The way they open. The way they close. The motivational kicker that appears in roughly the same position in roughly the same tone across thousands of different prompts.

You can read more about this in our AI writing patterns database.

The reason this matters is not aesthetic. It is functional. When your content pattern-matches against the AI writing fingerprint, readers have been trained to skip it. Not because they consciously recognize it as AI, but because it carries the same entropy as everything else they are scrolling past. It does not interrupt.

Eliminating these patterns requires more than better prompting. It requires a system that knows what you actually sound like, so it has something to hold the output against.

Why Bloomberry Is Not Just Another AI Writing Tool

There is a meaningful difference between a blank-page AI tool and a system that learns.

ChatGPT does not know what you have already posted. It does not know what you edited and why. It does not know that you tend to open with a specific kind of observation, or that you never use bullet points more than three items long, or that your audience responds better when you lead with a personal story rather than a claim.

Bloomberry is designed to learn all of that — from your actual writing, not a description of it. You can read more about how the voice learning pipeline works here.

The more you use it, the smarter it gets. Every edit you make to a generated post is a signal. Every post you select and publish becomes part of your voice memory. Over time, the system stops generating the average of all possible writing and starts generating something that is structurally closer to you.

That distinction matters as AI becomes the default starting point for content. If everyone is starting from the same blank-page model, and nobody is layering in persistent voice memory, then everyone is converging on the same output space. The differentiation is whoever has the most specific and well-trained voice layer on top.

Audience Before Product

One thing I talked about on the episode that I want to repeat here, because I think it is underrated:

Start building your audience before you have a finished product.

I know this sounds counterintuitive. But the value of an early audience is not just distribution. It is trust. And trust does not arrive the day you launch — it compounds over the months you spent posting about the thing you were building before it was ready to ship.

In an AI-saturated world, where generic content is everywhere and free, trust and repeated presence are becoming the durable layer. People need to have seen you think out loud, disagree with something, share what you are working on, before they are willing to pay attention when you ask them to try something.

That window of compound trust is not available to you if you wait until after the product is ready.

The First Paying Customer Came from One Comment

Here is a concrete proof point from the Bloomberry story.

On day two of launch, the Bloomberry LinkedIn account had three followers. I left a comment on someone else's post. That person clicked through to the profile, then clicked through to the site, signed up, and converted to a paying customer immediately.

There was no funnel. No ad. No automated outreach sequence. One comment, from an account with almost no presence, on a platform we had been on for about 48 hours.

The comment was thoughtful. It added something to the conversation. The person who converted was not just buying a tool — they were responding to a signal of genuine engagement.

That story has stuck with me because it is the clearest possible illustration of something I believe: founder-led, community-led distribution can create real demand at any stage, including zero. You do not need an audience to start. You need to show up and say something worth reading.

Why I Am Bootstrapping Before Raising

I get asked about this a lot, so I want to be clear about the reasoning.

I am not against venture capital. I have worked at companies that used it well. But there is a sequencing argument that I think a lot of early-stage founders miss.

VCs are not betting on your product. They are betting on your ability to grow fast enough to return their fund. That is a specific type of pressure, and it shapes every decision you make — including decisions about product quality, customer focus, and how you spend your time.

Right now, Bloomberry is in a phase where the most valuable thing I can do is stay close to user feedback, build the right product, and develop credibility organically. Taking institutional money before I have real product signal would mean trading that clarity for urgency I do not yet need.

The threshold I mentioned on the episode: I would not expect most VCs to take a serious meeting until you are at $2K to $5K MRR. Not because that number is magic, but because it represents proof that someone is willing to pay repeatedly — which is the minimum signal that a market exists.

The Bigger Point

Execution is getting easier. Anyone can generate a post in 30 seconds. Anyone can produce a video, a thread, a carousel, a newsletter.

What that means is that generic execution is no longer a differentiator. And the things that are getting more valuable — because they are harder to replicate — are authentic voice, consistent presence, memory of what you have said before, and judgment about what market conversations are worth entering.

ChatGPT can draft. It cannot tell you what you have already published. It cannot remember how you edited the last ten pieces. It cannot tell you when the conversation you are about to join is one where you actually have something to add, versus one where your comment will look indistinguishable from every other AI-assisted take on the topic.

Bloomberry is being built around the belief that those are the layers that matter. Voice. Memory. Distribution judgment. Authentic presence over time.

That is what I mean when I say voice is the last moat.

If you want to hear the full conversation — including the discussion on AI Sentence DNA, the Airwallex performance marketing experience, and the specifics on paid acquisition timing — the full episode with transcript is here.