Anthropic is a great case study in why uptime doesn’t matter. The service is so valuable that you can have one nine uptime and add $9bil ARR in 3 months.
Yep. Enthusiasts are cheap, picky, and have no loyalty. They’re extremely political and are the only type of customer who will actually switch. Plus it’s a tiny market. You might eek out $50mil revenue after a decade, if you’re lucky.
I built a payment processor and failed. Regulations aren’t the issue. The issue is customer psychology: once a customer has solved their problem, they never switch. The biggest misconception I had is that it’s a viable business model to start a business to “compete” with an existing one (cheaper, better tech, better UI, etc). That never works.
You have to either 1. Solve the problem for a new customer who hasn’t solved this problem. 2. Solve a totally different problem than your competition. or 3. Invent a completely different paradigm for approaching solving that problem.
A good case study is search. Nobody could compete with Google, until ChatGPT. Note that ChatGPT is not just “Google but better”, but instead does (3): it’s a different paradigm for answering your questions. Even though it’s much better at solving this problem, people still don’t switch from Google. Most of ChatGPT growth comes from (1): new customers, because ChatGPT usage is highest among young people who haven’t already solved their problem and aren’t sticky with Google.
You underestimate how sticky customers are out of habit. Cable news is now an inferior product for information retrieval, but it’s sticky because it’s already there, solving the problem, for that generation.
MCP blew up in 2024, before terminal agents (claude code) blew up in early 2025. The story isn’t “MCP was a fake marketing thing pushed on us”. It’s a story of how quickly the meta evolves. These frameworks are discovered!
> The story isn’t “MCP was a fake marketing thing pushed on us”. It’s a story of how quickly the meta evolves.
The original take was that "We need to make tools which an AI can hold, because they don't have fingers" (like a quick-switch on a CNC mill).
My $job has been generating code for MCP calls, because we found that MCP is not a good way to take actions from a model, because it is hard to make it scriptable.
It definitely does a good job of progressively filling a context window, but changing things is often multiple operations + a transactional commit (or rename) on success.
We went from using a model to "change this" vs "write me a reusable script to change this" & running it with the right auth tokens.
The best things about AI hypergrowth is the opportunities to discover of meta-frameworks and workflows. This is something Anthropic kills at (MCPs, Skills, Claude Code terminal agents).
These are discoveries of workflows. Some of them work some of them don’t. The ones that really click, they explode in popularity like OpenClaw.
Seems like OpenClaw is opening up a lot of breadth-based internet searches. Way more people are making way more scrapers, and I expect this to continue.
Yes. Stripe’s 2.9% fee minus interchange (interchange is variable, on average 1.9%) is higher than Mastercard or Visa’s take of .14%.
But it’s not so simple, because Stripe faces liability for merchant fraud. If you are high volume you negotiate IC+, where the plus is .1%-.4%.
The valuations price in expected growth as well as unit economics. Mastercard doesn’t have as much room to grow because cards already saturate consumer payments.
reply