magic got cheap... proof didn't

this morning i taped with an investor who spent seventeen years at google building assistants and cameras that understand the world. now he backs ai founders at the very beginning, from zurich.

he gave me a phrase i haven't been able to shake... commoditized magic.

magic used to be scarce. scarce meant expensive, and expensive meant margin. commodity was the opposite... cheap, everywhere, thin. ai broke that. things that looked impossible three years ago now cost about the same as a lunch, per month.

so if you're a founder, the hard question isn't can you build it. it's why can't your customer build it themselves this weekend.

and he thinks the bigger earthquake is still coming. within a couple of years agents start buying software, not just using it. an agent doesn't sit through your demo. it doesn't care about your logo or the three year contract. it tests you, checks the record, and switches the moment something better shows up.

think about what that does to the old playbook. lock in goes away. switching costs go away. what's left is proof.

he named three things that still hold. data nobody else has. expertise nobody else can copy. compute that's hard to get. i'd add a fourth, the one i keep seeing on my calls... people who already trust you.

half an hour later i caught up with an old friend who just moved to toronto. he works at a small shop that builds ai automation for companies. good team, real customers, and the same problem everyone has... they need more of the right conversations.

we didn't pitch each other. we just noticed that his customers keep needing what i do, and mine keep needing what he does. so we'll send each other founders. a partnership in ten minutes, because neither of us had to prove we were real.

that's the part an agent can't fake for you. it can compare your evals. it can't remember that you helped someone when you didn't have to.

so here's how i'd read the next two years. build the product like it'll be copied, because it will. build the proof like a machine will check it, because one will. and build the relationships like they're the only thing that compounds... because they are.

the founders who raise easy in a world of cheap magic won't be the ones with the best demo. they'll be the ones investors and buyers already know.

his whole argument is on the episode page if you want it from the source.

so if a machine could rebuild your product by monday... what would your customers still come back to you for?