The pace of company building is accelerating. The loop from idea to MVP keeps shrinking, and founders are prioritizing distribution earlier than ever. Everyone is talking about go-to-market as the great differentiator of the AI era — and rightly so. But the tech moat hasn’t disappeared. It’s being rewritten. And the rewrite is happening globally, across every sector, right now.
Warren Buffett gave us the original definition of a moat: a sustainable, durable competitive advantage. The tech version of that is unique technology that’s hard to replicate, creating a sure path to reaching and sustaining startup success (revenues, profitability, liquidation). For years, the focus was on the quality of the technology itself. In Israel especially, technical excellence has been a defining advantage often justifying premium outcomes, especially in cyber, where companies have historically demanded exceptional valuation-to-revenue multiples. But, as the barriers to building and shipping software collapse, the question is unavoidable: what happens to the Israeli advantage when anyone can build?
Many Israeli CEOs spend years in elite military units learning how to create elite technology under pressure. It’s no surprise that Israel leads the world in unicorns per capita and R&D spend per capita. But when writing code is no longer the bottleneck, the question shifts from “can you build?” or “how fast can you build?” to, “can you design a system for the age of AI?” That’s a different discipline entirely — one that rewards thinking and planning over coding.
AI-assisted development — vibe coding — has dramatically lowered the barrier to entry. More people than ever can create software from scratch, and they can ship faster than ever. The number of startups established per year is growing rapidly because of this. But, we believe the winners won’t be defined by who writes the best code. They’ll be defined by who designs the best systems: how the layers interact, how the product integrates with constantly evolving models, how adaptable the whole thing remains as the ground shifts underneath it. The architecture itself becomes the moat. Thinking matters more than typing. This doesn’t mean speed is irrelevant. It means the location of the advantage has moved. The strongest startups today aren’t just coding good products — they’re creating systems that get better as the underlying models get better. Their tech edge isn’t static. It compounds.
At NightDragon, four criteria shape how we evaluate whether a company is built for that compounding. First, a model-agnostic architecture — the ability to swap in better models instantly, without being locked into a single provider. Second, an automatic performance lift when frontier models improve, with no major rewrites required. Third, a real data flywheel, where product usage generates proprietary data that makes the system stronger over time. And fourth, an AI-native team — one that integrates AI across operations and workflows, not just inside the product.
The test we keep coming back to is simple. If leading models improve 10x over the next two years, will this company also become 10x better? If the answer is no, it’s structurally disadvantaged — and execution is unlikely to close that gap.
This explains a pattern we keep seeing as growth investors. We meet a steady stream of startups founded before the AI boom — many that raised at 2021 multiples, many doing $10M+ in revenue with real customers and solid businesses, by any pre-2023 standard. And yet, they’re stuck. Liquidity isn’t materializing, and investor interest has cooled. The reason isn’t the market. It’s that these companies were architected for a world where models didn’t improve so often. They’re now locked into a permanent cycle of catch-up — chasing AI rather than growing with it. We’re seeing that what starts as a technical limitation quickly becomes a strategic one, impacting velocity, margins, and long-term competitiveness, a gap which compounds over time.
To sum up, the next generation of winning startups will look different. The companies that win won’t necessarily create the best initial product. They’ll be the ones whose architecture lets them improve continuously, automatically, structurally, and at scale. Because in a world where the underlying technology improves exponentially, the only real moat outside of distribution is the ability to compound with it.
Dorin Baniel is a Partner at NightDragon.