There's a pattern in a lot of "growth has plateaued" stories lately, and I think the two usual culprits — AI Overviews killing SEO, and users vibe-coding their own tools — are actually the same root cause: cloneability.
AI Overviews hurt most when your product is easy to summarize. Users rebuild it themselves when it's easy to recreate. Both are the same problem wearing different hats — if the thing you sell can be explained in a sentence or rebuilt in a weekend, AI now does both for free.
Which leads to an uncomfortable question for anyone building right now: if a competent person with Claude can rebuild your core feature in an afternoon,
what exactly are you selling?
Here's the part I've been sitting with. Everyone's racing to build the clever AI feature - the generation, the summarisation, the agent. But the feature is the most cloneable part of the whole product. It's the first thing an LLM can reproduce. The genuinely hard-to-copy stuff turned out to be the boring layer underneath it:
Nobody vibe-codes that in a weekend. Not because it's intellectually hard, but because it's a hundred small correctness problems that only show up in production, with real money attached. The clever feature is a demo. The boring layer is the moat - precisely because it's tedious and unglamorous and everyone underestimates it.
I'll be honest about my bias: I build infrastructure in this space, so of course I'd argue infra is the moat. But I think the logic holds regardless of what I sell - and I'm not sure where the line is, which is why I'm posting.
So, for the people here: