When most apps are bitty throwaways and it takes an hour and $5 worth of tokens to throw up a marketing site, it becomes increasingly difficult to guess which new apps are worth using, which are giant security holes in the making due to blind vibe-coding, and which are malicious.

Sure, there were always sketchy fly by night app developers. But there used to be less fragmentation, such that it was easier for a worthwhile app to gain mindshare and you could actually learn what other people thought of it.

If it costs a day and $20 to vibe code a replacement (or a eeek and $100 in a corporate environment), I’d rather do that than risk the cost and inconvenience of installing something that will wipe my data or get me pwned.

I once saw a YouTube video of someone building a Starcraft-like game in under an hour using AI. When things like that become normal, I can totally see "why pay for it when I can just build it myself?" becoming the default mindset.

But I think this will naturally correct itself once companies start tracking AI costs seriously. If AI tools become standard in the workplace, the new performance metric could become something like "how few tokens did you consume to get this done?"

Once people start thinking in terms of token cost, rebuilding existing products from scratch every time starts looking pretty wasteful. At some point the token bill for repeatedly reinventing the wheel just exceeds the price of buying the proven product.