I've been in a couple startups that went Dynamo first and development velocity was a pale shadow of velocity with Postgres. When one of those startups dumped dynamo for Postgres velocity multiplied immediately. I'd estimate we were moving at around 1000% and the complete transition took less time than even I expected (about a month). Once the business matures, moving tables onto dynamo and wrapping them in a microservice makes a lot of sense. Dynamo does solve a lot of problems that become increasingly material as the business evolves.
Eventually, SQL's presence declines and transitions into an analytics system as narrower, but easier for ops, options proliferate.
I think this is out of date https://aws.amazon.com/about-aws/whats-new/2020/11/you-now-c...
1. DynamoDB is not as convenient. There are a bit too many dials to turn.
2. DynamoDB does not have a SQL facade on top.
3. DynamoDB is proprietary, I believe there's no OSS API equivalent if you want to migrate out.
4. DynamoDB was kind of expensive. But it has been a while since I last check the pricing page.
It's simply much better to start with PostgreSQL Aurora and move to a more scalable storage based on specific uses-cases later. For example: Cassandra, Elastic, Druid, or CockroachDB.