177. What Happened at Amazon and What Can Be Learned?

It was a bloody day at Amazon with 14,000 layoffs. This seems to have included almost their entire games division. I predicted that here on my substack in July, but can we prevent a repeat?

Yesterday I did my best to pull the curtain away from Microsoft to reveal The Wizard. But the moral of that story, like the original Wizard of Oz, is that the Wizard is just a man. A man with self interest and friends who have self-interest.

Now before I get into talking about the trees, I want to briefly talk about the forest. It will become easy to lose that high level perspective as I start to dissect Amazon, and particularly Amazon Studios. Yesterday I watched a video about how blacksmiths hundreds of years ago built iron tools and weapons far superior to what we make today, in their back yards. With essentially no knowledge of chemistry and only the beginnings of metallurgy. It’s not that we lost that technology, it’s just that we can make a million cheaper versions rapidly in factories at much lower price points. The end result is higher profits and lower quality (but higher quantity). High quality tools are labor intensive, and labor is so 20th Century.

As companies consolidate to form duopolies and oligopolies, and human labor is replaced by automation, all of society will become lower quality and more disposable. Even our jobs. Maybe especially our jobs, until the concept of “jobs” becomes historical and eventually mythological. Maybe a few extreme specialists (like myself) will hang around, until we age out or get caged as I described in the Microsoft paper.

“Blue” collar workers might be the last affected, which will flip our economies upside down.

Thus whatever will result in the lowest quality and most disposable product will be what we can reliably anticipate in all sectors with the possible exception of luxury goods. Our Masters don’t have to play by the rules they hand down to us. Every year, as food gets more expensive and human labor gets cheaper, brings us one year closer to the Soylent Green scenario that had a big impression on me when I saw it in 1973 at the local Fox cinema in Santa Monica. When billionaires fly their sexual partners into low Earth orbit while our SNAP food safety net is removed, the population reduction countdown clock I described in February becomes less fantastic.

Full Disclosure: I’ve been boycotting Amazon for at least 30 years since I observed them engaging in non competitive processes to build a monopoly by not paying state sales taxes. This ended up devastating the infrastructure in my native California. Massive pot holes and homelessness? The loss of small businesses? Even the hiring of ruthless traffic cops to terrorize the populace in order to restore State and local revenues? Amazon and Proposition 13 (which I talked about in my EVEOnline paper in 2013) were major contributing factors. Thus with the exception of the one story I’m about to tell, I have no direct exposure to Amazon.

The Call is Coming from Inside the House

In ~ 2017 I got a strange email from Amazon Game Studios. Or rather a new hire at Amazon Game Studios. This was pretty normal for me when I was still writing at Gamasutra. I had a massive readership and spie… er FOLLOWERS in every major studio. It was fair to say I knew almost everything that everyone was up to. The few remaining gaps were pretty easy to fill. When I’d be asked to sign an NDA, I’d chuckle and remind them that I was the only one in the room with any secrets. I just needed to know which of their many problems they needed help solving.

But this problem was very unusual. AGS had hired a “game economist”. While actual game economists were still extremely rare in 2017, Amazon certainly could afford any of the few that were out there. I’m not counting the army of Gameloft “game economists” that I described in my 2018 Game Economics Defined paper. Nonetheless, Amazon had hired a catfish with no relevant skills to what Amazon was building. Later this would be revealed as their New World MMO. That wasn’t unusual either as the noise to signal ratio for game economists was rapidly rocketing past 100 to 1. For major developers, getting an economist that could build the entire integrated metagame for an MMO would be like winning the lottery.

What made this call so unusual was that the call came from the catfish. He wanted to subcontract me to do his job so that he could keep his job. He claimed that his supervisor had actually authorized him to contact me for this purpose. The catfishes know who I am and who to call in an emergency, even if HR doesn’t.

I told them I’d be willing to help. Not out of desperation, but more from fascination. I had recently left Wargaming after a Catfish tried to extort me into doing something similar for him. But this situation with Amazon was so bizarre, that I was riveted due to morbid curiosity. Sadly, the scenario eventually fell through under the weight of what was likely multiple catfish inside AGS.

Thus while it was clear that whatever Amazon was spending so much money on was dead before it even started production (you can’t make a profitable MMO without a meta design), the extreme nature of this disfunction provides us with a lot of information about Amazon and AGS.

HR was Out of Control

I started writing about how the unscientific methods used in HR employee recruitment were destroying companies back in 2017. That was before it became terminally automated. If humans were screwing it up, it’s fair to assume they were programming AI to screw it up more. More in the sense that as technology accelerates a process you can expect the quality to decline even further as I explained at the start of this paper. This makes HR cheaper, much cheaper. So all the big players are moving to automate the HR process to save money. HR skilled labor is possibly the fastest disappearing specialty due to AI.

Now only key hires are vetted by humans. Even then, the list of people to vet will be AI generated. There is no one to question this because no one on staff is qualified to evaluate key hires. In regular industry a few catfish won’t torpedo your production chain due to redundancy. But game development is very specialized, and one key hire catfish might be game over. Two or more and it's almost certain. I still sometimes get vetted by humans, which makes me feel all warm inside. Simultaneously AI is getting better at identifying catfish, so the vetting lists are more likely to contain actual skilled labor. It’s still confused by LinkedIn profiles designed with Search Engine Optimization techniques that are used to trick AI. Nonetheless we are at that nexus where AI really is better at this than the average HR recruiter. Thus the massive layoffs in that sector.

Amazon is really into automation. It’s almost like mammals creep them out so they try to limit the number of humans running around in the company. They also do weird shit to make the humans know just how disposable they are, like making them pee in bottles instead of getting breaks. So of course there were likely massive numbers of catfish hired into AGS. A lot of money was flying around, so the people who were legit hires were probably really talented. They had to carry the weight of a dysfunctional studio while keeping their mouths shut. The actual humans in HR were likely in the extra “keep their mouths shut” zone. This reduced the chances that catfish would be replaced with proper hires. It was probably understood they were in the studio, and who they were. HR just covered it up to save themselves.

This sort of Self Interest which I described in detail two papers back is just a natural reaction to being placed in a high threat environment where your employer is perceived as disloyal to you. Handing you a pee bottle when you say your bladder is about to burst or telling you you can’t leave while a deadly tornado heads for your warehouse naturally reduces loyalty. Even for Amazon employees not in warehouses or tornado zones, this has to affect their perception of employer loyalty. The result is Low Morale which results in reduced teamwork, productivity, loyalty, reduced honesty, and can even spawn sabotage or commercial espionage.

If you make 100 tools in the time it used to take to make 1 tool, you don’t care if 10% are defective and they only last for a couple years. That’s where the money is. The problem is, that when it comes to teams, you are dealing with human supply chains. A supply chain is only as strong as its weakest link. Amazon’s business model specializes in creating weakest links at the lowest possible price point. This isn’t just a model, it’s an ideology.

Rumors have circulated widely about disfunction in AGS almost since its founding. Now that perhaps 99% of the employees there have been fired, I would imagine we are about to get a lot more intel about what went down there. It isn’t going to be pretty. I’m not shaming the legitimate employees there. If you put up with that nightmare scenario and kept your mouth shut, you probably deserve some sort of medal.

The Riddle of Wealth

Conan had “The Riddle of Steel”, which due to my origins is of great personal interest to me. But also due to my origins and unusual skill sets I’ve ended up working for and even living with some Extremely Wealthy People (EWP). They don’t follow the Riddle of Steel. They follow the Riddle of Wealth:

What is the one thing money can’t buy?

If you answered “Love”, sorry you didn’t win any prizes. People love money and will love you long time if you have it. The one thing you can’t buy with wealth, is Trust. The more money you have, the more self interested people will be interested in you. Likewise, the more money you have, the less you can trust those around you because they all want your money, not you. You might laugh at this Rich Person’s Problem, but if you think about it, it’s a really stressful life. My lack of self interest makes me super valuable to these sorts of people. That doesn’t mean they are loyal to me. People don’t become EWPs by being loyal to others.

So when Jeff Bezos wanted to create AGS, I assume he wanted it as a prestige branch of the company. Similar to the movie division that makes Lord of the Rings content. Being profitable wasn’t as high a priority as it normally would be. I think its safe to say Jeff is not a gamer. He could solve that problem with enough cash. But there is one problem you can’t solve with cash. Who to trust with this new vanity project?

Mike may not have been good about games, leadership, technology, science, hiring, consumer behavior, data, hardware, or empathy, but he certainly carried the ideological torch of the company. Most importantly, he was trusted by Jeff Bezos. Given that relationship, I find it odd that he left the company in 2022. Clearly that trust had its limits.

Mike’s appointment to this most critical role really set the tone for the whole studio. AGS was going to ship games like Amazon shipped books, Mike’s specialty. Input from those expensive veteran hires was largely ignored, especially if it didn’t match that “ship it faster, cheaper” philosophy.

This approach would cause many conflicts within the studio. While books and computer games may have both been entertainment and educational media, they are very different animals with extraordinarily different technological and scientific foundations. Yes the target audience in both cases is the human brain, but Amazon doesn’t make books, it just ships them. Shipping a book is not making product, it’s being a service provider. A computer game is a relatively complex product. A high budget MMO is such a fantastically complex product, especially on a scientific level, that its normal for people with 20 years of experience working on them to not understand how or why they work.

Frazzini was not aware of his mismatch to this task, and that lack of empathy/leadership meant that he was unteachable to the role by his employees. Due to his leadership role, it is fair to trace almost every problem below him… back to him.

Sure a lot of ignored industry veterans were employed at AGS. But how do you even fill key hires if you don’t know what a key hire is in a supply chain you are completely unfamiliar with? So even with the best on-paper veterans available, it seems likely that they were not in the correct roles, didn’t have sufficient authority to do their jobs, and there was no functional chain of command.

The Science of New Realities

I introduced the concept of New Realities in my 2018 The Physiology of Gaming paper. What does human physiology have to do with New Realities? It turns out, everything. Reality is subjective, and fantasy can become reality with enough immersion and affinity. The human brain is the consumer of MMO’s. How deep they go, and for how long, are the key metrics of success. Conversion (spending) is relatively easy to arrange. With sufficient neuroscience knowledge, you take a consumer pretty deep. So deep that you can actually kill them if you are careless about how the tech is deployed.

AI is starting to run into the same issues.

This is not the correct place to put your catfish. Or your “ship it faster, cheaper” guy. You can make a small house out of books. You can’t make a skyscraper with books. You can make a stool with books, to reach up higher. But you can’t get to the Moon with books. At some point you run into issues like gravity, weather, human physiology (humans need oxygen and temperature control), etc. More books doesn’t solve that problem, these issues require advanced scientific knowledge combined with sophisticated engineering.

An MMO is a much higher tech thing than a skyscraper. It could be argued that an MMO is even higher tech than a rocket ship since it took 30 years longer for us to create one. Without the requisite scientific and engineering skills, even a budget of a trillion dollars isn’t going to make it profitable. Actually, the more you spend, the harder it gets to be profitable. Thus this wasn’t a project that could be solved just with money.

Most MMOs are not profitable, and even with a lot of veterans, that could just end up making it harder to become profitable. This is a recurring theme on my substack because that’s one of my primary roles in the industry: to teach developers how to make profitable Games as a Service. Half of my most recent papers focus on various aspects of this (164, 167, and 168). I have at least 40 papers on this subject.

“More” and “Faster” are useless qualities if you don’t have the ability to apply the basic science of GaaS to your production/design chain. It works on shipping books but not making GaaS. Even movie making has a lot of science/theory involved, but despite being around for 100 years movies are much lower tech than GaaS. Directors on movie projects have some training but usually little science. The great directors are those that can use intuition to sort out the parts they don’t understand. This leads to a lot of volatility in regards to whether a movie will be a success or not, even with the same director.

You don’t see this volatility when science is applied. We don’t have an 80% collapse rate on skyscrapers. Once we kind of figured things out, even rockets rarely blew up. Whatever is going on at SpaceX seems to be a special case as they push the technology envelope.

Clearly the story of AGS was a tale of mismatched skills, ideology, leadership, and technology.

Building New Studios is Hard

When Wargaming went from 40 employees to 4000 employees while I was there, this was a rough time. When you hire that fast, all sort of organisational problems appear. Many people will just not be good fits or have proper qualifications. A super talented hire might fill the role of 5 other people and thus make those others redundant. Those 5 people might figure that out fast enough to put the trouble maker into the wood chipper.

Trying to make a big studio fast, to go after high prestige products, is a path fraught with peril. Making a great studio takes time, if it’s possible at all. Start with smaller more complex projects and then advance once your teams get to understanding their strengths, weaknesses, and natural chain of command. Western studios have a nasty habit of laying off key people after a successful product is released, then assuming they can do it again.

A top tier studio is a living breathing thing. Paying attention to its care and feeding is critical. The difference between a studio that can 10X its budget in revenue may only be a few key hires away from a studio that creates total losses. Even if you get everything perfect and preserve your team, you then run into the problem that your consumers change over time. So just because you made a similar product three times in a row, all successes, doesn’t mean a fourth will be successful. You could have done everything right, but now consumers just are fatigued and don’t want that product anymore. Data won’t predict that, because data tells you what just happened, not what’s going to happen. That requires science. Science supported by data is the ideal situation and that’s how scientific teams operate.

The content of this paper was easy to write, because I’ve been writing about it for years. Predicting a studio failure is easy, I do it regularly in writing so that my readers will know that with enough education these events are predictable and not mysterious. What isn’t easy is seeing our industry taking so many body blows. The layoffs are especially painful, and in most cases would have been unnecessary with a bit more forethought, planning, or education.

Some of these employees had to relocate for those jobs, some with family in tow. As an industry we don’t pay to send these people back where they came from. They are stranded. It’s a very unfair industry and unfortunately even the best talent often finds this out the hard way. I dream of a day when we treat our people better. I can’t say that day is getting closer, seeing how we are using science and technology to eliminate, not protect, people.

Thus I fear things will get worse before they get better. The tech industry divestiture that I predicted in July is in full swing now. I imagine most of my followers were hoping I was just talking shit. My long term followers know I do talk a lot of shit, but that shit happens. The worst part is probably the feeling of powerlessness. Knowing what’s coming but having no power to prevent it because the people who do have that power aren’t very good listeners.

Don’t forget to think about the forest.