Building has never been easier. That's why it's so hard.
July 22, 2026
What is left of competitive advantage?
We live in the best possible time to build startups. And somehow, it also feels like the hardest time of all.
Over the last four years, since around 2022, I gradually moved from launching side projects — wrappers on OpenAI, starting with GPT-3.5 — to more and more complex ones. You can see that path here: iworkedon.com/@dg.
The first product was an Excel formula generator: you wrote what you needed in natural language, and it returned the formula to apply. Trivial, today.
Back then, it felt like magic.
I did it for a precise reason: I wanted to escape the 9-to-5 and build something of my own, contribute to a larger vision, something worth suffering for.
Then, somehow, it happened.
November 2024: I was at the Vento bootcamp and joined Softgen, a product making $1.2k MRR. Lovable, for context, was around $40k MRR at the time.
I documented all of it on LinkedIn, and anyone who follows me already knows the story.
What happened next is still worth telling with a few numbers. Because they show how fast this market moves.
Lovable crossed $100M ARR about eight months after hitting its first million. By November 2025 it was at $200M, by January 2026 at $300M, and shortly after at $400M, with only 146 people. Today it is around half a billion in ARR, with a $6.6B valuation.
That same product was at $40k MRR when I joined Softgen.
Since the launch of Suna last year, the market has literally changed. Building software has never been easier.
Today everyone repeats the same line: models are no longer the moat. Coding is no longer a moat. Technology itself is no longer a moat, with rare exceptions.
If a technical user can paste your prompt into ChatGPT or Claude and get 80% of your output, you do not have a product. You have a wrapper.
And careful: starting as a wrapper is not a sin. Dropbox started as a layer on top of S3. Stripe started as a layer on top of a payments API. The problem is not beginning as a wrapper.
The problem is staying a wrapper forever.
So what makes a company competitive today?
For me, two fronts:
Knowledge: born from proprietary data, not publicly available, that improves with use.
Infrastructure: which requires serious capex.
Between those two extremes, in a bell curve, sits everything else: wrappers of wrappers of wrappers.
After the Softgen exit, and after leaving Kortix — official on July 31 — the question I keep asking myself is the same:
What do I want to do?
What can I do?
Where can I create value?
And this is where I come back to the title.
We live in a world that is extremely easy to build in, on the software side, if we can still call it that. And at the same time, extremely complex.
Today, four years into my entrepreneurial path, I sometimes feel empty, almost consumed.
Especially by what I see around me: everyone is doing exactly the same thing, changing only the UI.
So where is the value?
If I started a new company today, in which sector would I build it?
How do you cultivate knowledge (because it is easier said than done)
The short answer I give myself is this: knowledge is not bought. It is accumulated.
And it accumulates in only one place: where your product goes so deep into how someone works that it becomes part of the process.
Harvey, the AI for legal work, does not win because its model is smarter. It wins because every brief, every review, every firm-specific precedent makes the next answer better for that firm, and model providers never see that data.
That is a moat. Everything else is just a starting advantage.
In practice, cultivating knowledge means three things for me:
1. Choose a vertical domain boring and specific enough to scare generalists, then become obsessive about it.
2. Build a loop where every user interaction leaves behind data that the next competitor cannot replicate from zero.
3. Have distribution that model providers do not have.
It is no coincidence that the next winners will probably not even present themselves as "AI companies."
They will be knowledge companies, workflow companies, infrastructure companies. AI will be an ingredient, not the whole product.
One thing I would definitely do today is this: study.
Not get pulled into FOMO.
And figure out how to cultivate that knowledge for an edge not in a month, but in a year.
It is an extremely complex and delicate process. But I know it may be the only way to work on something different, something with real impact.
We cannot expect to work on the same thing for three years, given how fast everything evolves.
At the same time, we should cultivate a ten-year vision, knowing the inner soul of the product will probably change twenty times along the way.
And this is where it all holds together
The speed of the market — the same speed that took a product from $40k to half a billion ARR in a little over a year — is not the enemy.
The enemy is believing that building fast is enough, when building fast is now something everyone can do.
Resilience matters more than ever. But not the resilience of someone who keeps their head down on the same thing forever.
The resilience of someone who can hold uncertainty long enough for knowledge to settle, while everything around them changes twenty times.
The easiest time to build is also the hardest time to build something that lasts.