@josevalim As we move more and more responsibilities onto AI agents, where do you see humans sitting in the loop? I've found myself caring deeper about systems, networking, databases and how the whole puzzle fits together but models might catch up there as well
@MarcJBrooker@thorstenball@copyconstruct I understand the analogy but one is deterministic, other is not. How do you blindly trust a system that is non deterministic for mission critical things?
@GergelyOrosz While i understand your point, and I do agree with the general sentiment, we got to remember that we live in a tech bubble. Microsoft has the enterprise market outside of tech cornered.
@jerzydejm@eeuoss A good design for 1k qps is not a good design for 100qps, and a good design for 100k qps is not a good design for 1k qps.
Design for 10x scale, and have a solution in your back pocket for 100x scale.
@JustJake > This works for standalone Postgres services and for Postgres HA clusters. On a cluster, the primary is upgraded first and every replica is rebuilt from it on the new major; the cluster stays paused (no automatic failover) until every member is back on the same major.
@JustJake might be a comprehension issue on my end, but downtime between major version upgrades is a bit of bummer no?
For context, I decided to not use railway for my side project due to this specific reason