I have spent twelve years on AWS. IAM, Organizations, VPC, ECS, Lambda, RDS, KMS — the mental model is deep enough that I stopped noticing it was a model at all. It was just how cloud works.
This week I enrolled in the Google Cloud Gen AI Academy APAC Edition, and the track I picked ends with deploying a Gemini-backed Streamlit application to Cloud Run. I have not done it yet. I am writing this first, deliberately.
Here is why: once you have solved something, you reconstruct the experience as though it were obvious. The genuinely interesting information — which assumptions you carried in, which ones were load-bearing, which ones quietly failed — is only available before. So this post is a set of predictions I am putting on record, from reading the documentation, before I touch the console. The follow-up will mark them right or wrong.
Everything technical below comes from Google's documentation, not from my own hands. That distinction matters and I have tried to keep it visible throughout.
The map I built from the docs
