I would say that serverless does a bad job handling databases, especially Postgres, since they are designed around a small number of high-volume TCP connections while serverless causes lots of minimally-used sockets to be spawned putting heavy pressure on Postgres' fork-per-socket design.
I also had very bad experiences with AWS Lambda being very disconnected from how you would typically structure and deploy a backend service; dead endpoints wouldn't actually be deleted without special care being taken, so you could accidentally have code in production that still works using endpoints you thought you removed and no longer have the source for (actual issue in another startup I joined), and of course the API endpoint structure (URLs, HTTP methods allowed, payload provided) were dictated by Lambda, not the needs of your frontend or backend code. This implicit scaffolding in Lambda also makes it very difficult to do integration tests locally.
So AnyCloud lets you structure your backend code how you want, so local testing is about the same as how it would behave in production, and it handles not just autoscaling like Lambda but also the deploy/monitor/autoscale/rollback needs for you, and since it produces instances in your cloud more similar to the kinds of clients DBs expected, its much harder to break postgres.
I also had very bad experiences with AWS Lambda being very disconnected from how you would typically structure and deploy a backend service; dead endpoints wouldn't actually be deleted without special care being taken, so you could accidentally have code in production that still works using endpoints you thought you removed and no longer have the source for (actual issue in another startup I joined), and of course the API endpoint structure (URLs, HTTP methods allowed, payload provided) were dictated by Lambda, not the needs of your frontend or backend code. This implicit scaffolding in Lambda also makes it very difficult to do integration tests locally.
So AnyCloud lets you structure your backend code how you want, so local testing is about the same as how it would behave in production, and it handles not just autoscaling like Lambda but also the deploy/monitor/autoscale/rollback needs for you, and since it produces instances in your cloud more similar to the kinds of clients DBs expected, its much harder to break postgres.