It’s time to go bigger

TL;DR

• 1

  AI changes the economics of software, not just the speed: Theo says the scary part is not that 100 hours of work becomes 15, but that builders now need to rethink what is worth attempting when engineering labor gets dramatically cheaper.

• 2

  The cloud is the historical analogy: Just as AWS made experimentation cheap by removing the need to pre-buy servers and predict traffic, AI reduces the cost of hiring and building enough that old assumptions about team structure, product scope, and delivery no longer fit.

• 3

  Old startups had to win narrow verticals: Competing with AWS or Salesforce used to mean picking one slice and going deep, because trying to match their full feature range was a "suicide mission" for a startup team.

• 4

  Horizontal products now look more viable: Theo's thesis is that a "shitty all-in-one" that is functional across many categories can suddenly make sense, especially if users can extend the parts where they need more depth.

• 5

  Lakebed is his proof-of-concept: He is building Lakebed as a "shitty cloud for shitty apps" with framework, runtime, hosting, database, auth, and deployment baked together so small internal or one-off apps stop dying in dashboard hell.

• 6

  The bottleneck moved from writing code to assembling systems: Theo's own pain was that deployment, OAuth, databases, inference providers, and preview environments still took hours even after app code got cheap, which pushed him from building glue tools like Shoe toward rebuilding the stack itself.