When I arrived I was taking my time to understand the culture, company, needs, etc. before making any suggestions. One example, they did Agile with sprints and I think KANBAN works better. But, I didn’t see it as an important issue to spend time on.
So the Google guy comes in and from day 1 began making suggestions for big changes to both process and the software architecture. He often started by saying, “At Google we…”
I was let go, in part, because the CEO thought I was not contributing to certain technical discussions. I told him I thought the proposed change wouldn’t bring any value to potential customers because it was a purely internal architectural change. We had a lot of actual customer facing work to do and this was a distraction.
So… a guy with only 5 years of professional experience all at Google won over the 26 year old CEO more than someone with 20 years of experience across multiple companies and having built a very similar product just a year before.
The new developers managed to convince the leadership that we needed "better" processes, and other technologies. The result of this was rewiting the system into more micro services than developers in different languages and multiple frameworks. The new processes ensured that teams didn't talk together because each team should be independent. The startup failed hard and barely survived even after major cuts
Large companies also tend to have sociopaths become the VPs, who only care about their career progression and not the actual company direction.
“Grok” is from Heinlein’s Stranger in a Strange Land. It was a super important word to me as a kid. I’ve recently started hearing it again used by people who never read the book and find it interesting.
Do you remember where you first heard this word? I assume you heard it as it’s such a memorable word and difficult to misspell.
I heard about it from graybeards and HN, hence why there might be a resurgence now that HN is mainstream.
In 95%+ cases that’s a pretty terrible idea
Everyone expects them to be building Kubernetes, Terraform, and other useless bullshit that a startup with a total critical traffic of 10QPS absolutely doesn't need.
We've also had ex-Facebookers back in the day who would throw out perfectly fine coding processes and get everyone to use Phabricator. Bloated software that took many engineering hours to design processes around, and came with features that no one with Atlassian would ever downgrade to use. It was like coding with PHPBB.
2. Terraform isn't a Google product. It's by Hashicorp
Also, do you really need it when you're a startup running 3 EC2 clusters with "potentially tons of customers"? No.
I've never worked at Google, but if you want a clean set of commits to implement a feature (think Linux kernel/subsystem, git, etc), Github makes it unnecessarily hard to do that. The best they have is squash and merge which basically makes one mega commit and a merge commit that references a branch with a single commit.
With squash and merge you get chunky, meaningful commits that (with discipline) are tagged to a specific issue/feature so it's really easy to understand the history.
I think ‘stg’ with a GitHub publish/stack sync option would be a winner.
Feeling rather uncomfortable as Devil's advocate in this case but I was testing Oracle Cloud Infrastructure as a cost-saving replacement for Google Cloud Platform recently and they are worlds apart.
GCP is a cloud service where you can issue a command or click on a web panel to get your resources up. Oracle, on the other hand, will set up multiple meetings with various "specialists" to teach you how a cloud should be used. Unproductive meeting after meeting after meeting and no work gets done.
I felt like an unpaid QA engineer when going over all the unpleasantries of their Terraform provider with them. Documentation is garbage and no variables are are explained[1]? No problem, we've got 5 specialists who will train you how you should use the cloud.
Here's a generous free tier so you can test everything. Oh, want to set up resources using it? Raise a support ticket. Want GPU VMs? No problem, request the hardware for a specific region days in advance and we hope it'll work. Oh, not working? Please share your screen and open a support ticket. Want a second VM instance or a node pool for your cluster? Request limit increase. Want more RAM? Request limit increase. Don't worry, we have a team granting these very quickly.
Absolute shitshow. Oracle Cloud is as frustrating as Digital Ocean when you try to treat it as a cloud and not a VPS vendor. GCP more or less "just works".
[1] https://registry.terraform.io/providers/oracle/oci/latest/do...
I've always been impressed by the software quality at Google. My biggest complaint is... I'm kind of a test fanatic. In my mind, 100% code coverage is a great starting point for software that might run OK in production and delight customers. But at Google, people are so smart that they can keep a million line system completely bug free just out of sheer willpower. Every CL they write or review, they know exactly what it will break, and just tell the author to not break that. The end result is that there are 0 tests for the most critical software around and it all works out somehow! Ever visit google.com only to see "google.com is not responding"? Nope! You don't need tests when you're a super genius.
Overall, I think Google builds really good software internally. Everything I used while I was there, I long for in the "real world". Github's code review system just makes me cry every time I use it, even 6 years after I left Google. Critique was so much better! It looked nicer, the UI told you who actually needed to approve the CL, there was a dashboard that told you what work you actually needed to do, etc. Great software. Monarch is another favorite of mine. It is light years ahead of anything you can get in the open source world. When I left Google I begrudgingly selected Prometheus and Grafana as my "hey is production down?" system of choice? And every day they make me sad because they're so fucking primitive compared to what I could use at Google. I remember a long discussion thread between me and an InfluxDB founder on HN about whether or not it's possible to write certain queries. The query in question was a query that I had on a public dashboard at Google, that ran in milliseconds over a stream of 1 million plus cardinality. "There's no way a computer can do that efficiently," he told me. I don't know, I've had a computer do it efficiently before with the literal first query I typed in. And that's the market leader in time series data, probably two decades behind some unloved internal tool at Google. Their software is top fucking notch. They just have no idea that it's even something that can be sold. (The people at Google selling software rewrite or acquire sub-par alternatives. At least there's Cloud Spanner. That's a good database!)
I think I'm inclined to agree you about management at the highest level. Like I said before, Google has no idea they can sell their internal tools and make more money than most startups rewriting the tool from scratch could ever dream of making. They choose to go on multi-year tangents that yield nothing. Google Buzz? Google+? You got slapped on the wrist by the FTC for that shit, it didn't sell a single ad, and then the market leader you were trying to kill died anyway because everyone got bored with social media. Dumb. Let's talk about GSuite. As a card-carrying Google fanboi, I never believe that anyone else could make good webmail or a calendar. The startup I work at got acquired recently and we had to switch from Gsuite or actually is it Google Workspace For Your Domain or something now, anyway we had to switch to Outlook. Outlook on the web is great! The web design is beautiful, it's super fast, and it has a million more features than Google's stuff. Thinking back, I'm not sure anyone has typed a single line of code into any of those products in over a decade. They're just a dinosaur waiting to die! (Microsoft has it's own problems, though. They write a bunch of great code, and then ... just specify dumb shit. Like at work, my name is "Rockway, Jonathan". Who the fuck calls anyone by their last name? It's dehumanizing. It makes me not want to go to work. But that's how some intern coded up Active Directory Nineteen Ninety Four edition, and so it persists to this day).
I don't know where I was going with this comment, but yeah, the tech behemoths of yesteryear area dead. Remember "FAANG"? Facebook, completely dead, fired all their employees. Amazon? I guess they still exist to take CFOs out for golf in exchange for paying $1,000,000 for a $2000 server and $20 of electricity, and send me the best fire hazards from the Shenzhen market overnight. Apple? Honestly they're doing OK for a big company. Sure, they're working on VR goggles instead of a Clock app that works, but you know... marketing. Netflix? Dead. Google? Top Three In Cloud. Amazing. You know what they say, if you can't be second best, at least you can be third best.
This post was long and incoherent and I am now quite inebriated. I hope it was moderately enjoyable to read. I'm not a crazy person, I promise. It makes me sad to see what the MBAs and CFOs did to Google. Great company, and I'll never stop loving it.
The tools you use are culture-defining.
And the managers... Oh, the managers... They just act for the sake of their own promotion even if that means damaging someone's else career or the company in the long term. And will complain about things not being done the "Google Way", even if the proposed Google Way failed multiple times in that context (startups and scaleups, in my case). But what's really shocking is how they have no interpersonal skills, to the point of making you constantly question yourself: how, why, did this person ended up in a management track in a supposedly Y-career company? How not only did they got there but also promoted multiple times for this role?
Google, as a company, is as cool as Oracle nowadays.