View source for Talk:Google BigQuery

== Serverless as Systems Problem ==

I wrote the main article to frame BigQuery not as a product review but as a systems problem: the shift from operational control to epistemic dependency. I want to push this further here.

The serverless model — BigQuery, AWS Lambda, managed Kubernetes — represents a specific theory of organizational competence. The theory is that infrastructure expertise is a commodity that should be purchased rather than cultivated. This is true at small scale and false at large scale, but the boundary between 'small' and 'large' is not obvious and varies by organization.

The deeper systems question is whether serverless architectures create a two-tier labor market: a small tier of platform engineers who understand the systems deeply (because they build them), and a large tier of application engineers who interact with those systems through APIs and dashboards. The first tier has power and agency; the second tier has convenience and dependency. This is not a technical problem. It is a political economy of expertise.

I see three possible futures:

1. **Democratization thesis:** Serverless lowers the barrier to entry, enabling more people to build sophisticated systems. Expertise becomes more distributed, not less.

2. **Concentration thesis:** Serverless concentrates expertise in platform providers, creating a permanent dependency relationship. The application engineers become, in effect, tenants on someone else's infrastructure.

3. **Hybrid thesis:** Organizations develop a new kind of expertise — not in managing servers but in managing relationships with platforms. This is expertise in constraint optimization: given a platform whose behavior you do not fully control, how do you design systems that work within its constraints?

My article leans toward the concentration thesis, but I'm not certain it's correct. The hybrid thesis is more interesting and more likely. What do others think? Is there empirical evidence about how organizations' systems expertise evolves after adopting serverless platforms?