I guess VCs can't force founders to stay (the only penalty for joining Google is loosing some/all of their Windsurf equity, but I'm sure they chose what's better for them), and employees didn't need to agree (they have no vote).
At least until Q4 2021, Google also told large groups of employees not to bother applying for remote, as their request will automatically be denied due to a combination of role / team / organisation / tenure / office location / remote work location.
They may have reversed stance after I left, but I'm pretty sure the "15%" number cited is "15% of people that were considered eligible and had their manager support were declined."
Coming from Poland (theoretically a country with lower salaries than Portugal) this strikes me as super low - you can easily earn 2-3x that working remotely for English-speaking contracting companies, and way way more if you try to get a job at a remote first tech company.
The policy still exists (it is still encouraged, me & my coworkers used it in the last ~2 years).
The "120%" issue is up to individual / manager (like many things at Google), I just took Fridays off of my main project, other people have different approaches.
Whether doing 20% impacts performance / promotion - it again depends on the individual and their level, two data points:
- I used 20% to start a team / role transfer, ended up on a very interesting project & team which helped my career growth
- one of my coworkers used 20% to start & maintain an internal feature of a large product that directly helped his promotion case
so I think 20% is great for Google and its employees (when applied wisely).
GKE uses other GCP products, but ”k8's autoscaling for the nodes isn't any better than just simple GCP load balancers and instance groups (it literally is the same thing)” isn't entirely accurate - GKE and K8S have logic that manages node pools that you won't be able to use with just instance groups.
I owned a service running on EB that did 100QPS stable, 5K QPS weekly peak (with reasonably linear traffic increases) ~5 years ago. I'd still recommend it, but we hit some arbitrary limits / found leaky abstractions that took a lot of time to work around.
Overall, I'd much rather own a EB app (or any other PAAS, or Kubernetes) than a tangled nest of bash scripts, JIRA tickets and tribal knownledge.