[shadowgovt]

That has ceased to be true in vast tracts of the company. Cloud, for example, is chasing feature parity with its competitors and engineers are told to implement things because AWS or Azure have them, not because they organically grew out of the architecture of the existing system. In fact, quite a few features annealed poorly onto the existing architecture (which users of Google Cloud may have memory of).

[ayberk]

Huh, citation? :) I'm not saying you're wrong, but as a GCP engineer I have never seen us (well, my org) implementing something because of "AWS or Azure has it".

[shadowgovt]

Memory's a little fuzzy, but I'd put IAM and workspaces in that category.

IAM, in particular, was a huge undertaking in jamming fine-grained access rights onto existing resources where none such existed before, and it was pretty much marching orders from above: "Potential clients can't migrate off AWS because AWS has this and we don't." And it caused more than its fair share of "Why is this API suddenly throwing errors" tickets from existing users who were accustom to the pre-IAM permissions model.

ETA: Re-reading my initial statement, it was over-broad. There is room in Cloud for bottom-up engineering and product design. However, especially relative to the rest of the company (where Google is an industry leader, not entering a market already heavily dominated by an elephant), Cloud spends a lot of its time chasing "table-stakes" features to enable new customers to be on-boarded who can't subscribe to Cloud because they can't migrate their existing flow off AWS without X Y or Z analogous feature available in GCP.

[ak217]

I have been interviewed multiple times by GCP product managers seeking to understand how my team used specific AWS services that Google lacked at the time to learn how to implement Google's competing service.

I would be more concerned if your org wasn't doing this. AWS does it all the time too. The parent post is wrong to paint this as a negative. And GCP has some unquestionably industry leading products too (BigQuery, GKE, Spanner, AI/ML services)

[benlivengood]

Online disk resizing was the one I tracked as it slowly trickled through alpha/beta. To be fair, pretty much every virtualization product had online resizing before GCE, so it could have been a matter of prioritization.

Whatever the heck Vertex AI is compared against the Sagemaker+Ground Truth pile that AWS has. By no means is any of it groundbreaking from either company, just piling open source software and buckets behind UIs, but it seems GCP is doing catch-up there.

[brunoTbear]

Yeah I think this is rather disconnected from my experience building products inside of GCP.

I'd be curious where you think we could have done better on the annealing with specifics :)

[shadowgovt]

IAM rollout was a chore (and introduced a whole layer of abstraction that users who didn't need fine-grained resource management now needed to be aware of / care about) and workspaces interacted with the existing billing infra in a messy fashion, if memory serves. On the plus side, that was years ago and AFAICT Cloud learned the relevant lessons from those stumbles.