Hacker News Reader

Comment by bilater | Hacker News Reader

"I predict one day Amazon will fail. Amazon will go bankrupt," Jeff Bezos.

One of the most surprising things in my brief experience at Amazon was how much of a shitshow it was. The two-pizza rule and self-sustainability for each team led to huge overlaps, with teams doing the same thing.

With such a huge organization, you had to go through 15 different stakeholders to get a single thing done, and there was an ingrained middle management whose only function was to connect you to the right person.

Just figuring out who was responsible for what and how to get things done was a challenge in itself.

Despite all this, Amazon still succeeds, and their process of PR/FAQ, leadership principles, and one-pagers is one of the best I've ever seen.

But I wonder if at some point, like with any philosophy and just like Bezos predicted, it will become too much and the whole thing will cave in on itself.

View on HN

To give Amazon and Bezos credit, their 14 leadership principles give a pretty balanced framework for decision making. Customer obsession and working backwards to help focus on what's important, disagree and commit to resolve differences that root in intuition or assumptions, bias for action to minimize analysis paralysis, and etc. It's amazing that Amazon can sustain its size for so long. But yeah, eventually it is people who enforces culture, and eventually an empire falls.

The LPs are bullshit. They are basically templates that the people in power can (ab)use to force their narrative. Want others to shut up and do the job? "Disagree and Commit". Don't want to spend money on employee's well-being? "Frugality". All else? "Customer Obsession". And what does "Are Right, A Lot" even mean?

It means that hierarchy matters more than otherwise, because the higher-ups "are right" more by LP decree than by meritous consensus.

Also they have a dumb rule to dog food everything...which sounds great but that includes basic software like their own Github.

They had this monstrosity called Code Commit that I saw recently is getting canned. Good riddance. What an ugly pos.

> Also they have a dumb rule to dog food everything

This is false. Otherwise I'd be interested in any reliable sources.

As far as I can tell this is some sort of urban myth since AWS first stepped onto the scene.

I believe it's based on Steve Yegge's original microservices rant: https://gist.github.com/chitchcock/1281611

I don't think that's myth, though I've heard it criticized variously as outdated, or just one person's take.

There's nothing in Yegge's rant that implies Amazon must dog food AWS. AWS as we know it, was, if anything, a byproduct of the efforts described by the rant.

Amazon certainly uses AWS but I think it's foolish to think they use every product (they obviously do not) nor that they must. For the foundational pieces -- EC2, SNS, S3 -- sure, of course. They spent a long time migrating workloads.

I don't expect that with every new service they launch that they have already widely adopted it by then. Maybe after their customers widely adopt it, if ever.

> There's nothing in Yegge's rant that implies Amazon must dog food AWS

Really? From the link:

> "The Golden Rule of platforms is that you Eat Your Own Dogfood."

and from Yegge's interpretation of Bezos' mandate:

> "All [mandated internal APIs to be used for all interservice communication], without exception, must be designed from the ground up to be externalizable. That is to say, the team must plan and design to be able to expose the interface to developers in the outside world. No exceptions."

Of course Amazon doesn't necessarily use every single AWS feature internally; I don't think dogfooding implies building all your features for internal customers first, and then selling them. Rather, it implies refraining from selling crappy reimplementations/duplications of internal tools.

They subject us to meet with them as customers on Chime. This is despite everyone hating the product. CodeCommit was annoying because nobody uses it, and you had to touch adjacent projects like CodeStar to do something like trigger on a GitHub commit.

> With such a huge organization, you had to go through 15 different stakeholders to get a single thing done, and there was an ingrained middle management whose only function was to connect you to the right person.

If you're able to recall, can you list who these 15 stakeholders were? Are they like legal department approval, etc?

Reply on news.ycombinator.com