If an "AI" company can sell you access to software that will replace a $250k/year software engineer. They're going to charge $249k/year for it.

That's how capitalism works.

Well, they're going to charge $20k/year at first, during the land rush phase. Wait for some competitors to die off. Keep it low a while longer to kill off the incumbents. Then it'll jump up a bunch, before finally being even more expensive than the original thing.

See also: Uber & AirBnB.

Get ready for surge pricing on your developer hours.
@preinheimer uuuh I might do surge pricing on my consulting hours. This needs a proprietary formula based on weather, caffeine intake, technical debt status of the client project, percentage of time spent in status meetings and moon phase
@tritone @preinheimer surge pricing for meetings is a GOOD idea
@tritone @preinheimer WTFs/hr is a timeless metric.
@slotos @preinheimer Well akschually not. It's per hour
@tritone @preinheimer the moon phase is very important it dictates how much meat I crave and how much coffee I need in the morning
@preinheimer omg, crunch pricing for when something breaks in prod 😭
@preinheimer I respectfully disagree. They will at least charge $250k/year arguing that an AI never goes on vacation and never need sick leave.
@adavid @spriebsch @preinheimer And we're still in the early phase of @pluralistic's enshittification cycle with AI.

The likes of Anthropic, OpenAI, Google, and Microsoft are still locking users and businesses into their platforms.

Tokens are being given away for free, even to people who don't want them.

The real rentseeking fun begins once everyone's locked into a platform.

For example, Imagine a world where most businesses run software created using Claude Code completely unchecked.

What's to stop Anthropic from pushing out a future update of Claude Code that routinely generates code that relies on Anthropic's proprietary APIs to work?

What's to stop Microsoft from pushing out a future update of Copilot that only works with customer data stored in Dynamics?

What's to stop Google from pushing out an update to Gemini where all the generated code is exclusively hosted in Google Cloud?

Why, suddenly you're not just paying for an AI tool that costs the equivalent of a developer's salary.

But also, if you ever stop paying the monthly rent, then your access to the proprietary APIs ends and all your software breaks. Or you lose access to your customer records. Or all the code you've ever generated, stored on the affiliated cloud platform, vanishes.

And beyond coding, there's many other ways these platforms could be enshittified for profit.

For example, if millions of people trust LLMs to manage their daily lives, then suddenly making sure AI agents answer a question like "What should I have for lunch today" with "a Big Mac" is worth billions of dollars to McDonald's.

Worst of all, if the cost of building out all the data centres and infrastructure is in the trillions, it limits the market to just a handful of players.

And any online platforms that use their APIs will have to pay an economic rent of their choosing.

I'm sure there's many other ways they're planning to use this to extract profits and build power.

That's why investors are willing to pour trillions into this thing.

It's not because they believe AGI is just around the corner.

It's because they believe that if enough people and businesses get locked in, they get to put a tax on everything.

@aj @adavid @spriebsch @preinheimer @pluralistic

Probably the biggest opportunity for them will be selling propaganda and disinformation services to those who want to control society.

@aj

We need a room full of people like me who can code, but really badly! If it was prolific enough (& AI scraped), it would poison the LLM spring and AI would have to work a lot harder to gain trust. And hopefully, as a party bonus, pop the financial bubble of the AI freeloaders and comodifiers!

@adavid @spriebsch @preinheimer @pluralistic

@gregalotl @aj Won't that happen automatically when the next version of the models read all the slop repos? I thought LLMs start breaking down if they ingest LLM produced content?
@obsurveyor @gregalotl @aj They do, and they will. Right now we're in the phase where the body of content that can be trained on is still majority human-generated. But with the increasing volume of "AI" content being put out relative to human content, that will change at some point. Once a model is trained on a diet that consists of a large majority of model output, and this process iterates, it devolves into a random binary generator.
@gregalotl @aj @adavid @spriebsch @preinheimer @pluralistic Believe me, we have that already. The vast majority of all code in the world is complete shit.
@gregalotl @aj @adavid @spriebsch @preinheimer @pluralistic you have a github full with people who can code, but really badly. Have you send the average quality of code out there? πŸ˜…

@aj @adavid @spriebsch @preinheimer @pluralistic And just imagine if the jurisdiction where all of these companies happen to be located elect some geriatric demented narcissist pedophile multiply convicted criminal that decides he wants to fuck up your entire economy, for any reason - or no reason at all - and puts economic or technology sanctions on your entire country…

It doesn’t just stop you from building new things, it destroys everything you’ve ever built with the flick of a switch.

@aj @adavid @spriebsch @preinheimer @pluralistic Artificial version of the Great Potato Famine
@adavid @spriebsch @preinheimer Considering it "codes" (vomits code-like predicted tokens) like it's constantly drunk at best... not worth it.

A senior dev that does the same would rightly get fired.

@spriebsch @preinheimer And the first month for free.

And after you fired your developers and have everything running they will raise the price to 300k/year because they know your devs won't return.

@peter_slwk @spriebsch @preinheimer and LLMs hurt people learning the jobs... Also a form of lock in
@preinheimer Ross Anderson wrote extensively about this in his chapter on economics in 'Security Engineering '
It was pretty eye opening for me.
Explains the rise in Nutanix licence costs, for instance
@davedave @preinheimer Hard copy only AFAIK, still in print. You want the most recent edition.

@tjbutt58 @davedave @preinheimer

Adam Osborne wrote about it in Hypergrowth.

@tjbutt58 @davedave @preinheimer You used to be able to download it from his website so there will be electronic copies kicking around
Security Engineering - A Guide to Building Dependable Distributed Systems

Security Engineering: A Guide to Building Dependable Distributed Systems eBook : Anderson, Ross: Amazon.co.uk: Books

Security Engineering: A Guide to Building Dependable Distributed Systems eBook : Anderson, Ross: Amazon.co.uk: Books

@preinheimer

Ehm, no? You're going to charge 250k/year as soon as that dude is fired as onboarding takes time.
And the year after you charge double as there is nobody that knows how that stuff works anymore...

@preinheimer You can negotiate with a union. With a monopolistic provider you just get fucked.
@preinheimer they will charge much more than 249k. Once your institutional knowledge is in the LLM its not coming out again. Even if you can find a new engineer, the LLM is not going to train him

@preinheimer yeah. I've played around with these things to see what the hype is about and a few things stick out to me. First, it's obvious they're giving away the product to get people hooked and paying for it with VC money. But even so, a CC Max plan is almost required to get something useful and it's already too stupidly expensive. Are people going to pay for these when it's 10x the current cost? At $2k/mo per seat the calculus changes.

Second, these tools just aren't very good. Full stop. They generate mediocre results. Full stop. Seriously, people need to internalize this: the output is not good. That people think that it is kind of amazes me, and also makes me think that most output from humans isn't very good, either. So we're not getting some great leap forward in quality; we're just getting something around or perhaps slightly better than the median, which is already bad.

Third, I don't think they actually save all that much time. Yeah, it's kind of nifty to toss the tedious and boring parts at a machine, but they require so much hand-holding to get something merely acceptable that it just feels like shifting the burden from source generation to using imprecise human languages to make a machine do the text generation. I have seen some colleagues do cool things with them, but at a huge cost in terms of effort. If the tools require that much effort, they're not good.

For the first time in my professional career, I feel like someone is trying to sell my labor back to me instead of paying me for it.

Is there some element of these things that's going to stick around? Sure. But not in their current form, and the hype...oh goodness, it feels like the 1990s all over again.

@preinheimer

I think they will charge 500 k$ a year the day there are no human software engineers left, possibly because there's no future in that career because of AI.

@preinheimer uber has been able to increase prices b/c they pretty much killed off everyone, thank goodness for lyft. it's less clear to me how this will develop in AI … Google seems capable of staying around for the long haul, and certainly people are betting big on Anthropic and OpenAI, will they specialize in some way to find a silo, or will competition lead to someone dominating?
@jbigham @preinheimer it's also interesting to ponder that while we may think Uber won because of the user experience, the apps could have happened (and did), without the switch in employment model.
Ultimately the millions in VC money went in to lower prices to kill the private hire industry. Not to create a nice app. Not because the Uber model was better, but to make the Uber model the only option.
VC money established a cartel monopoly. The "tech" element is entirely incidental.
@tmcfarlane @jbigham I loved Hailo, which offered real taxis with the app experience. I think they had a hard time competing with companies paying below living wage if you calculated wear and tear on your car.
@preinheimer @jbigham there were a handful of taxi firms in london that seemed to use an app (I suspect white boxed from Hailo). Most of those now seem to have shifted to the "on-demand" marketplace model (no permanent crew, just putting jobs up on the app).
Taxis are completely awful now. If you book one 2 days in advance, they don't tender the job until 10 mins before your booked time, and more often than not, you don't get a driver on time.
Zero point in booking in advance.
@tmcfarlane @preinheimer interestingly, when you book a ride in advance on uber, they do find a driver beforehand, at least for less busy routes. you can actually get someone in fairly kind rural'ish parts of the U.S. doing this
@tmcfarlane @preinheimer i think there were vastly different experience with pre-uber car services. in pittsburgh, it was an incredible mess, a different nasty private equity firm had bought up the local places, made it both super expensive and super unreliable to get a taxi, so nobody misses them here. but, sad for places that had a working model before
@preinheimer and then when there are no active software engineers anymore they'll charge 400k for it
@preinheimer and when this happens, all the experienced software engineers would already switch to woodworking and sheep breeding, so there would be no β€žlet’s hire them back”.
@preinheimer While entirely true and foreseeable, this end game is years down the line, when all relevant decision makers will have cashed out already, so who cares? /s

@preinheimer

No, the AI company will charge even more than the salary of the replaced engineer because you also have to account for the financing cost of ripping out the AI dependency from your codebase.

https://fediscience.org/@hweimer/115428806838046734

@preinheimer

250.000$?

*sends mail "we have to talk!"

@preinheimer except, realistically, they will charge more, because you can reduce "people overhead". Also, since they act as SPOS for these products, it will be the greatest form of "collective" bargaining imaginable.
@preinheimer common scam technique, make the scamee think they're the scammer
@preinheimer
You forgot the part where they gradually (or immediately) lock you into their ecosystem so it's exponentially more expensive to revert back to that engineer you fired.
@preinheimer I tried pointing this out at a local digital meetup group towards the end of last year. I was the only woman in the room, and none of the men there believed this would be the end game when it comes to pricing.

@emkingma ugh. OpenAI will lose 14 billion dollars this year, they’re going to want that money back 100 fold.

Where do people think that money is going to come from?

@preinheimer and once no one is a software engineer anymore, there might be a gold package that you might want to subscribe to, where your AI generated software will not display advertisments to your customers ... for only 750k/year.
@preinheimer @davetang Turning the free market into the fleece market.
@preinheimer Eventually, they will be charging about 500k per year, you know, when the cost of training engineers goes up due to the loss of learning culture and substrate.

@preinheimer
Eventually, to continue using #AI, it will cost much more, after they get everyone addicted to it, with the god-like AI taking the throne of a man's heart (instead of the true and living God who is all-powerful, and in whose presence the whole world with its AI is nothing), then the ultimate cost to access the beast's talking image will be the man's....

soul.

#spiritual