We have taken a very aggressive approach. We are divesting our professional services business and have rebuilt from scratch on AI. The stack is Lovable + Supabase on Vellum. We have moved a lot of our process and data into context documents that we manage inside Vellum. We use three main LLMs plus an AI built for doing Bayesian math. This allows us to innovate very quickly. Given the speed at which we are moving, fungible credits are the right pricing model for us. We had to invest a lot of effort into getting the design of the credits right so that credits provide roughly the same value across different agents and use cases. We are modelling gross profit at 70%, lower than a traditional SaaS business, as we spend a lot on compute and even as per token costs come down with the amount of reasoning we consume our overall compute costs should be stable. We are self funding so we do not have to justify the lower margins to investors. As an angel investor the lower margins of companies that are AI native are acceptable as long as they are > 60%.