Data is your only moat
Substack ● Covered by 3 sources
An investor essay argues AI moats aren't models, they're data collected through easy adoption or deep enterprise integration. Why it matters: it explains why coding agents exploded while slide-generation and support bots face wildly different competitive futures.
Based on reporting by Substack — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The framing here is a 2x2 grid: problem complexity on one axis, ease of adoption on the other. Plot coding agents, customer support bots, slide generators, and SRE tools on it, and a pattern emerges that has nothing to do with how hard the underlying task actually is.
Take coding. It's genuinely a hard problem, yet Cursor-style tools improved faster than almost anything else in AI. The reason isn't cleverness — it's that any developer could switch IDEs in five minutes with zero approval needed, generate code dozens or hundreds of times a day, and every accepted or rejected suggestion became training data. That fast, high-volume feedback loop turned into a flywheel. Slide generation sits in the same "easy to solve" bucket technically, but lacks that tight feedback signal, so progress there has crawled.
The piece's sharper point is that easy adoption is a trap as much as an advantage. If a startup can win a market in five minutes with no procurement process, so can OpenAI, Google, or Anthropic — and they'll do it with more users and more data. ChatGPT Health gets cited as exactly this kind of land-grab. Meanwhile the
My take — AI-written commentary, not fact-checked reporting
I'd bet on the boring, hard-to-integrate stuff every time — nobody builds a durable business selling something a frontier lab can clone over a long weekend. The real tell here is that everyone chasing quick consumer AI wins is racing Google and OpenAI to the bottom, while the companies quietly wiring themselves into messy enterprise systems are the ones nobody will bother replacing in five years.
Read more about this at: Substack