Pizza, calzones and Stripe

The Stripe CEO compared software to pizza this week. He runs the financial plumbing for half the internet, and he was not being funny. Your SaaS business, it turns out, has the shelf life of a calzone.

This week, three of the most powerful people in tech 𝗱𝗲𝗰𝗹𝗮𝗿𝗲𝗱 𝘁𝗵𝗲 𝗦𝗮𝗮𝗦 𝗲𝗿𝗮, 𝗮𝘀 𝘄𝗲 𝗸𝗻𝗼𝘄 𝗶𝘁, 𝗼𝘃𝗲𝗿.

Patrick Collison said software should be "cooked right then and there, like pizza" — custom, on demand, bespoke.

𝗡𝗮𝘃𝗮𝗹 𝗥𝗮𝘃𝗶𝗸𝗮𝗻𝘁 𝗽𝗼𝘀𝘁𝗲𝗱 𝗳𝗶𝘃𝗲 𝘄𝗼𝗿𝗱𝘀 𝘁𝗵𝗮𝘁 𝗴𝗼𝘁 𝟯.𝟭 𝗺𝗶𝗹𝗹𝗶𝗼𝗻 𝘃𝗶𝗲𝘄𝘀: "Pure software is rapidly becoming un-investable." That's the guy who spent a decade telling everyone to invest in pure software.

Michael T., the Cursor CEO, quietly disclosed that AI agents now write 35% of his company's own code.

None of them were talking to each other. They were describing the same thing: the death of a business model.

Marc Andreessen wrote "𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗶𝘀 𝗲𝗮𝘁𝗶𝗻𝗴 𝘁𝗵𝗲 𝘄𝗼𝗿𝗹𝗱" 𝗶𝗻 𝟮𝟬𝟭𝟭. His firm has deployed ~ $42 billion on that thesis. He has not updated the essay. He has, however, updated the portfolio.

𝗧𝗵𝗲 𝗼𝗹𝗱 𝗦𝗮𝗮𝗦 𝗱𝗲𝗮𝗹 𝘄𝗮𝘀 𝗴𝗲𝗻𝘂𝗶𝗻𝗲𝗹𝘆 𝗯𝗲𝗮𝘂𝘁𝗶𝗳𝘂𝗹.
Write code once, sell it to 10,000 companies, collect $200/month per seat, forever. Fixed costs, effectively infinite upside, zero marginal cost on delivery.

𝗖𝗼𝗹𝗹𝗶𝘀𝗼𝗻'𝘀 𝗽𝗶𝘇𝘇𝗮 𝗺𝗲𝘁𝗮𝗽𝗵𝗼𝗿 𝗶𝘀 𝘀𝗵𝗮𝗿𝗽𝗲𝗿 𝘁𝗵𝗮𝗻 𝗶𝘁 𝘀𝗼𝘂𝗻𝗱𝘀. When software gets generated on demand, inference is your ingredients. You cannot price like it's a download when serving a customer costs something real.

When your product builds your product, what exactly are you selling? 𝗖𝘂𝗿𝘀𝗼𝗿 𝗶𝘀 𝘀𝗲𝗹𝗹𝗶𝗻𝗴 𝘁𝗵𝗲 𝗺𝗮𝗰𝗵𝗶𝗻𝗲, 𝗻𝗼𝘁 𝘄𝗵𝗮𝘁 𝘁𝗵𝗲 𝗺𝗮𝗰𝗵𝗶𝗻𝗲 𝗺𝗮𝗸𝗲𝘀. Worth pausing on that.

𝗪𝗵𝗲𝗻 𝗡𝗮𝘃𝗮𝗹 𝘀𝗮𝘆𝘀 𝗽𝘂𝗿𝗲 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗶𝘀 𝘂𝗻-𝗶𝗻𝘃𝗲𝘀𝘁𝗮𝗯𝗹𝗲, he means: your moat was code nobody else had. If an agent can replicate your core product on a Tuesday afternoon, that was never a moat. That was a head start, and head starts expire.

𝗧𝗵𝗲 𝘄𝗿𝗶𝘁𝗲-𝗱𝗼𝘄𝗻𝘀 𝗵𝗮𝘃𝗲𝗻'𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱 𝘆𝗲𝘁. Trillions in SaaS valuations still say "build once, sell forever." They will.

𝗝𝗼𝗯𝘀 𝗰𝗮𝗹𝗹𝗲𝗱 𝘁𝗵𝗲 𝗰𝗼𝗺𝗽𝘂𝘁𝗲𝗿 𝗮 𝗯𝗶𝗰𝘆𝗰𝗹𝗲 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗺𝗶𝗻𝗱. Naval's update: maybe now we have a motorcycle. You still need to ride it.

𝗧𝗵𝗲 𝗹𝗮𝘀𝘁 𝗱𝗲𝗳𝗲𝗻𝘀𝗶𝗯𝗹𝗲 𝗺𝗼𝗮𝘁 is something AI cannot ship overnight: whether your users would actually miss you. Not "find a replacement." Miss you.

The 2011 thesis was "software is eating the world." 𝗧𝗵𝗲 𝘄𝗼𝗿𝗹𝗱 𝗯𝗶𝘁𝗲𝘀 𝗯𝗮𝗰𝗸.

Build something people love, not something that reheats well.
Never been cheaper or faster to do that.

𝗧𝗵𝗲 𝗰𝗮𝗹𝘇𝗼𝗻𝗲, 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗿𝗲𝗰𝗼𝗿𝗱, 𝗮𝗹𝘀𝗼 𝘁𝗵𝗼𝘂𝗴𝗵𝘁 𝗶𝘁 𝘄𝗮𝘀 𝗽𝗶𝘇𝘇𝗮.
The window to become the pizzeria has just never been this wide (h/t Elena Verna)

Also, Marc Andreessen hasn't updated the essay. Should he?