The enshittification of the tech industry is reaching a fever pitch, and it’s all thanks to AI. Not what AI can do, but what people seem to think it can do.
It’s really hard to get a job in tech right now. Surely our industry has gone through its ups and downs and hiring waxes and wanes alongside interest rates and how sweaty the many perpetually damp men in the Silicon Valley VC scene are on a given day. But things are getting rough out there, friends.
I always hesitate to use the overused term “perfect storm,” but maybe this is one. Interest rates are high, VCs are less emboldened (unless your pitch has something to do with AI or the blockchain), and there is this massive shared hallucination that generative AI is going to allow a company to do everything it currently does, at the same or better levels of quality, with a quarter of the staff.
So certain are the grasping capitalists that this narrative is true that they’re firing people in droves. All of these talented people suddenly entering the job market alongside this assuredly false belief that they won’t be needed has created an imbalance in the market. Too many tech workers, too few tech jobs.
But what’s really happening here is enshittification.
Cory Doctorow coined the term “enshittification” for this phenomenon where a company locks in a consumer audience with unrealistically good prices (typically subsidized by VC), then uses the leverage of a massive and committed consumer audience to lock in suppliers. It is at this point where all the benefits begin to be shaved away. First the prices go up for consumers, then the payouts go down for suppliers. Amazon is so good at this that they’re in multiple federal lawsuits about it.
The enshittification of the tech industry is similar. Employers like Google and Netflix realized how much leverage they could get with great software talent, so they pampered the shit out of these kids with Michelin Star chefs and on-site laundry. Eventually their products became the defaults (the market share of Google as a search engine is still over 90%, an absurd winning streak for a search service that’s getting gradually worse).
Once the audience was locked in and the cash was free-flowing, they could start taking the screws to the employees. Maybe there’s still a barista in every kitchen, but it’s “come back to the office five days a week” and “your weekly target should be about 60 hours of work” (said Sergey Brin recently, a man who just completed his slow-motion transformation into a sociopathic ghoul).
It is unequivocally the AI narrative that is fueling this enshittification. Not what AI can do, but what leaders think AI can do. This narrative has taken hold for two important reasons: first, there is a seed of truth in it; second, it is exactly what corporate tech leaders want to hear.
The seed of truth is that generative AI in the programming space has advanced to a truly usable level in just the last few months. My own work is surely accelerated by the help of Copilot, which mainly types faster than I do, and which can convert large swaths of JSON into a similarly large swath of Golang structs faster than I can… But it’s not really solving any problems. Copilot dramatically reduces the length of the “search StackOverflow, identify the accepted solution, copy and paste the code” loop.
Tools like Cursor and Windsurf might be even more effective, but ultimately AI can only patch together fragments of stuff it’s already seen, which has a really high degree of utility when building prototypes, but can’t solve any of the problems that only become problems when you actually succeed at something and face real competition.
To the extent that the “StackOverflow copy/paste loop” is at least half of the job, it’s very helpful! But it can’t replace me, because it can’t understand what PMs are really asking for, or what customers might actually want but don’t know how to ask for.
But the real cutting edge of this AI narrative is the second bit: it’s the story that these lonely, grasping, soulless, fragile male egos desperately want to tell. Mark Zuckerberg was probably as worked up as a nun in a cucumber patch when he got to tell his board of directors that he’ll be able to lay off thousands of pampered nerds and still produce the same output.
It’s a great story. It somehow survives scrutiny, at least from those with no hands-on experience. Some of the facts line up enough to suggest that it could work out. But… It won’t. It can’t.
Amazon is similarly using AI as an excuse to “flatten the organization,” which is a business euphemism for firing capable managers and forcing the rest to manage 30+ people each. Google tried that once and it was a spectacular failure, so I doubt AI is going to magically fix it. If you think it’s bad reporting to a hurried, burned-out, frantic manager, try reporting to ChatGPT. AI may seem patient but what it really is is relentless and confidently wrong a lot of the time.
You know what else is relentless and confidently wrong a lot of the time? Ego-drunk sociopaths. It’s like if every single engineer at Amazon suddenly reported directly to Jeff Bezos, a man who seems mainly concerned with how often he can get his shirt off to make sure you know how much hair he has under there. Gross. You’re gross, Jeffrey.
The light at the end of the tunnel is that this mass hallucination is just that: a hallucination. Maybe AI will fill in the gaps enough that juggernauts like Amazon can continue adding endpoints to their AWS APIs and only burn out 1,000 engineers instead of 10,000, but mark my words: the gravy train will jump the tracks eventually.
Just as the dot com bubble burst, and as the blockchain proved to be useful only to buy drugs and gamble, this AI “dream” will evaporate and companies will be stuck wondering why all of the products they are building seem to be regressing to some uninspiring mean.
Keep your heads up, friends.