The majority of Architects I've brushed against in the last decade have been exclusively an additional marketing force for AWS with very little ability to objectively reason from first principles or assist with knowledge of anything outside of the AWS ecosystem.
Totally genuine question as someone coming from a background in mechanical engineering, who now does a lot of software-y things: what do you consider to be "first principles" in computing? Are you talking about first principles in terms of project development processes (i.e how information systems evolve) or code (i.e. bits and bytes, where microservices might not be the answer because the network connections are guaranteed to have a fundamental latency that, along with the amount of data required, means you can't meet requirements because [..])?
Not the parent but I will offer you my opinion after I laughed at their comment because it reminded me of many of the engineers at my current work:
"First principles" in this context is not just reaching for the nearest/newest AWS service that seems to do what you want. ie. if you want a webserver, don't just run magic lambdas that abstract everything away from you, go and provision a vps and install your dependencies on it so you fully understand and control how the stack works.
None of the 12 engineers at my current company had written a test that actually wrote rows to a database when I arrived, they didn't know how. In fact none of them knew anything about how to run our databases on their own dev machines at all, they all shared one RDS PG instance.
Again, "first principles" here is: running a process, connecting to a service, asserting the whole thing did what you expected. To do that effectively you have to actually control the whole system, and to do that you have to know how the whole system works together. A lot of developers now just delegate the provisioning and maintenance of everything to a provider, even in development setups.
First principles to me is the basic foundations of an idea.
Abstractions are used but; any given abstraction is not owned by a single vendor and are based on agreed upon standardised abstractions as they relate to to operating systems and databases etc.
I think I can name an example, which is how I am approaching datalisp; https://sr.ht/~ilmu/tala.saman/ the first principles bit for my context refers to: fixing a standard encoding that is sufficient for the task, determining inputs and outputs of model, setting a general guideline for expressivity of model (space of conflict-free programs), sketching interaction modes and user interfaces.
These are the first principles in question, now from there your architecture is mostly determined, you can fall back on known patterns and decades of research. The axioms are the architecture.
