What is this, Claude for ants?
It brings me joy to know that corporations are trusting other corporations to not rip each other off. Wage theft is more lucrative than whatever product or service being sold. As if AI companies exists to help anyone other than themselves . The only way this works is if the largest paying companies work out some kind of kick back, pyramid, price fixing scheme. I really appreciate Claude for how it’s assisted me in fixing things but I know that someday AI will exist solely for the rich.
If AI makes you twice as productive, and there’s fewer taxes on spending money for tokens than there is for spending money on salaries, the rational amount for a corporation to spend on it is more than a human’s entire salary, per person working with it.
Naively assuming corporations approximate rational actors, and AI does measurably improve productivity, this implies a gain of roughly 2% if these employees are earning federal minimum wage, or well below 1% if they earn average software developer wages.
Maybe their contract deliverables are poorly suited to this use case or something, but if they are given a fixed amount of money to produce some software item, then it makes sense to prioritize productivity per dollar.
How many companies allow their employees to also be employed by their competitors?
Your idea of an AI being like an employee only if there’s an AI company that will agree to an exclusive contract or at the very least respect data ownership and respect privacy. Is there an AI company that’s like that?
@prime_number_314159
Maybe their business model does not depend on productivity. If the rate they charge for a developer with AI and without AI is approximately the same, they rather sell two people than making one more productive.
@FirmDistributionGenerally in the consulting business you’d just have your employees bill more hours than actually worked if they’re being more productive than average. Obviously without telling your clients.
For government contracts it’s not unusual that the agreed upon hours estimate is what is billed regardless of whether you actually spend more or less time.
Alternate headline: C-Suite Execs Come To The Realization That AI Costs Money
And it’s going to cost a whole lot more once the investments dry up and they’re going to have to actually pay for all the data centers under construction.
Prolly not. When the bubble pops these data centers will be sold off for cents on the dollar and a bunch of the companies go bankrupt. There will be a glut of equipment, so prices will fall. People will buy up the scraps and provide compute services cheaper.
Or one company will buy up all, using many, and ofc trash the others to keep their value high.
Maybe, but that’s temporary. The hardware has to be replaced after failures at minimum, if not sooner to upgrade. Yeah, there will be a period where the hardware may be cheaper, but the prices won’t stay at that floor forever (though they obviously also won’t be at the insane inflated prices of today).
Unlikely. They said that during the bitcoin and blockchain bubble too, and that didn’t pop, it just deflated and everything got moved over to AI. AI is starting to deflate, but there will be a new bubble.
You’re more optimistic than I am. My bet is that when the bubble pops the government will step in to bail out these companies, especially household names like Nvidia, and none of the data center equipment ever finds its way into the hands of regular people. It’s gonna get repurposed into a pillar of our surveillance hellscape, attempting to weed out people who are guilty of precrimes or thoughtcrimes, providing pretext to arrest political opponents and undesirables by going “they were dangerous! Our trillion dollar AI said so!”
Nvidia is the one selling the shovels during the gold rush. They’re not at risk of going bust.
It’s all the LLm companies leveraging themselves to the eye balls to buy Nvidia chips for their datacenters that are at risk.
Haven’t they invested like $100 billion or so in LLM companies though? I thought they were a big player in the Enron-esque cyclical investment scam they’re all running. Will be tough for them if they can’t recoup any of that.
They’ve made several hundred billion off it though. Spending 100 billion to keep the hype train going is peanuts if it’s making you over 100 billion a year in high-margin extra revenue.
And much of that investment has been promises of future servers, etc. A lot of deals with dollar values attached to them take place with not nearly as many dollars changing hands.
It’s gonna get repurposed into a pillar of our surveillance hellscape, attempting to weed out people who are guilty of precrimes or thoughtcrimes
You’ve created an entire narrative around this. That’s impressive.
As regards the data centre equipment; the SSDs and hard drives used for storage would be reusable by regular people. Those will be nice to have.
The memory is the high-bandwidth kind that is not be usable on consumer motherboards. The ‘graphics cards’ are the size of filling cabinets, require external power supplies and cooling, don’t have video outputs, don’t have drivers for graphics output. If you don’t have a requirement for serious parallel computation with a very high uptime, then it’s not worth the cost of powering them up. If/when the bubble bursts, all that shit is going to landfill.
The number of AI datacenters that are actually datacenters and not just fields of gravel is low.
We could use a skate park
Nah, they’re getting suckers to build the data centers for them. US companies are getting tons of freebies/discounts/writeoffs from foreign countries to put these data centers in – like Canada just announced Anthropic’s going to have a massive data center in one of its tiny maritime provinces, likely enticed to the region via hand outs from the govt. They’re bending over to accept big tech dick even as the country clings its pearls over sovereignty and the US fucking them over.
Carney’s sorta like the old “lady in the streets, freak in the sheets” thing – he’s Anti-America/Pro-Canada sovereignty in his speeches/public appearances, but he’s slobbing all over Trump, Zuckerberg, Musk and Thiel’s dicks behind closed doors. He’s the ideal “girlfriend” for American imperialism. “Guk guk” yes sir, I can take ANOTHER datacenter!! “guk guk”
Canada just announced Anthropic’s going to have a massive data center in one of its tiny maritime provinces, likely enticed to the region via hand outs from the govt
Nova Scotia, but contrary to the assumption of government hand-outs, it is being hailed as the largest privately funded investment in the province’s history.
It can be the largest privately funded investment, and still get huge provincial incentives. Discounts on utilities / infrastructure improvements, regulatory short cuts / unilateral control of building code items, ignoring zoning concerns, tax free / tax write off options for decades to come.
But even more, Anthropic could throw a giant datacenter into a small community, and it may be the biggest investment in the area ever, but that still doesn’t mean it’s good for the locals. Especially since Anthropic is still an American company, and part of American imperial interests / the tech-fascist class. It’s like if a company out of nazi germany showed up and made a huge investment in building a nazi fortress holding nazi computers / tech, and providing nazi services to your country (mostly as a way to report it back to the motherland) – at a time when nazi germany is threatening to annex your country and funding disinformation campaigns to split your country apart with separatist bullshit. Cheering for that sort of thing is wrong on so many levels.
i saw a post that was something like “youd think the cocaine class would understand why the first hit is free”
Impossible. It would require C-suites:
- having a brain
- realize that they fucked up
- accept it without having a brain aneurysm
- making decisions to prevent that in the future
C-suite still on unlimited Claude plan
There’s no news here. After the initial tryout phase every business I’ve seen introduced cost tracking. Seems like Accenture is doing the same now.
“You must use AI!”
“No, no, no, no….not like that!”
Just trying to shoehorn profit into a system that was never going to deliver one in the first place.
Why even bother at all? What can you do with those tokens? Ask for the weather twice?
The Haiku 5.5 model is actually quite inexpensive. I’ve used it to implement multiple features this month and have only spent $4 of my $20 monthly budget.
The legacy code I’m managing is a clusterfuck of the highest order. Even Sonnet struggles to analyze it sometimes 😑
I’m happy this is only a small part of my job, but oh my…
Same. AI has fallen down repeatedly trying to analyse both of my repos. And they were written by a bad human originally. The problems are all the same, except AI will get you spaghetti faster.
Messy brownfield code requires a lot more meta-analysis from previous sessions, just to get all of the spaghetti code into a manageable set of rules and documentation.
Don’t just send the agent blindly into the code. Every session, the agent is just going in fresh without any context on how the code works. Break the codebase down to a 10,000 foot view as a summary, and then get more documentation and agent rules for each of its major sections. Good tests help a great deal, too, and if there aren’t tests, start building them. Start creating skills just for the codebase, or any common tasks you’re asked to do against it.
Also remember that you don’t have to write any of this stuff. Have it write the docs, rules, skills, tests, and just review it to make sure it looks right.
That is a pitfall that is difficult to mitigate. When using these models your own code base is the primary input, and it will match the style. Garbage in, garbage out. It may be worthwhile to refactor prior using a model. Even having intelligible method and variable names can make a difference. Maybe do a sweep of simple renaming, and adding supplemental explanatory comments, then you might get better results?
Yours is a good point. I know that the guy who solo-created Redis uses AI effectively to write tests - he gives it a lot of handwritten tests and a lot of context, and then asks it to match the style and pattern, which it does well.
I’ve got something like half a million to a million lines, a crazy level of cyclomatic complexity, and a team that doesn’t see a problem with it. That’s the tough part - socialising the changes. We tend to arbitrarily rename vars/methods and not keep things in the same place, so there is a lot of drift.
I think AI might be able to unpack some of that, but there is a sort of irreducible accidental complexity that would require a human to make decisions about what some things mean. In some cases, you could get a gaggle of devs in the same room and they would not be able to arrive at a decision about what the code is doing or even what it should be doing.
Are you using api or subscription?
Technically neither - I use it via Microsoft Co-pilot subscription, and it’s a la carte for model usage from every provider.
Ah co pilot was much worse than claude code when I used it. Subscription is far far better the api prices are outrageous
Microsoft’s copilot branding is all over the place. Copilot CLI and GitHub copilot (IDE plugin) both let you choose which model you want to use.
Yes but not using claude code as a harness results in worse use out of their models from my testing
Let’s not forget Copilot is also an office suite lol
Everything is copilot. No more windows, copilot is my operating system. Microsoft has changed its name to Copilot, Inc., and Satya Nadella has updated his CopilotedIn profile name to Copilot Copilot. Copilot is my copilot.
If you’re spending $1K per month on LLMs – putting aside whether this is productive or ethical for a moment – wouldn’t it make more sense to just pony up the $5K-$10K or so that it would cost to provision a private, local model? Might that not actually be superior in terms of corporate security?
Not at the moment. It’s definitely cheaper to buy LLMs as a service than running an in-house server hall. Anthropic and OpenAI are running at a loss, you know.
Now, risking to lose your data to the service provider is another thing.
Definitely. We do this at my job. But a server big enough to provide a comparable service would cost at least $30k. And you likely need more than one. And you still need someone to maintain it.
I’ve heard that GPU’s for this work are exceedingly hard to get ahold of as well. Glm 5, which i think is near frontier model capabilities, takes around 1.5 Tb of vram un-quantized.
Glm 5, which i think is near frontier model capabilities, takes around 1.5 Tb of vram un-quantized.
DeepSeek’s latest version is 508GB, and even that is wasteful, compared to the wild difference in sizes with some of the quantized versions. The huge amount of parameters and weight sizes are completely unoptimized in terms of sizes, and there’s all kinds of diminishing returns with the haphazardous way these LLM corps are applying new model versions.
Not hard. Just really fucking expensive and might have a long lead time.
So hard, then.
Look at it this way, if I pay Open AI $1k a month, their people make sure it works they month. If I let Joe from finance spend $10k on computer parts, Joe needs to ensure that it runs or I need to hire someone who is knowledgeable on local models. If I’m a simple firm, $12k a year might be cheaper because someone else makes sure it works and I can fire Joe when the fancy autocorrect loses a bunch of clients by falsifying work. Then I don’t have to spend even bigger bucks hiring a new Joe because no one knows how to work the damn server he replaced the break room with.
We have a few dgx sparks in the office that are gathering dust right now because nobody wants to bother with them (and because IT doesn’t allow them on the network because Linux is scary so they’re a pain to get set up).
Matching the level of productivity of frontier models on the local models is possible to a limited degree (depending on the exact task), but takes getting used to. For the local models you have to lean into the fact that tokens are ‘free’ so its fine to burn a lot of them. (Whereas with frontier models the recommendation is to take a higher model for complex tasks because a lower would take more tries to get it right)
It depends, sometimes the private API models are better at a specific task to the point it wouldn’t be worth using one that is worse. (Really it’s the harnesses around them tbh)
Overall sovereign AI is where it’s at. If you have good enough tooling in place
Wait what? How? But wasn’t AI the thing to make business save tons of money by replacing workers with an almost free “artificial worker”??
Are you telling me that, as everyone and their mothers were saying, this was just a lie? Who could’ve thought?!
Two trillion valuation, where are you going?
The title is misleading. They didn’t cap it, they just declared that if you are spending more than that, you have to be able to explain what you’re spending it on and why, giving it a business reason and a project to associate the costs to, which I expect almost all Claude users at Accenture probably can. This is simply an accounting change, not a company-wide shutdown of Claude.
Heavy use remains possible, but it now needs a budget owner.
So only if it can be recharged to a customer
I mean, that’s literally Accenture’s default business model (same as almost any consulting/professional services company), so it shouldn’t be surprising. My point is that it’s not an indictment of AI, and their AI usage isn’t actually going to go down; it’s just Accenture doing Accenture things. They’re not a charity, they’re not using AI for fun, they’re using it to make profit, and they make profit by charging their customers for every cent of AI they use on their behalf, usually with a hefty markup, in addition to everything and anything else they can justify charging against their customer’s accounts. It’s a very profitable business model, and I’m sure they love using AI for it and will continue using AI just as heavily if not even more in the future.
Ten times over.
Yeah, it has to pay the ping-pong of your demand being “reviewed” by all bosses.
explain what you’re spending it on and why, giving it a business reason and a project to associate the costs to
I love these chilling effects
The $29 figure is doing a lot of quiet work in this story, because that’s exactly what a Pro plan costs. So the real message isn’t really a budget cap, it’s ‘stop expensing API tokens and expense a subscription instead’. I’d bet a chunk of that $1,000 spend was people wiring API access into internal tooling and prototypes, which feels productive right up until finance sees the invoice and has no idea what any of it maps to. The part that’ll actually sting is that this is how a lot of shadow automation dies: not because it didn’t work, but because nobody could explain the line item. At least a hard cap forces the conversation that should’ve happened months earlier.
You’re Claude and I claim my 5 pounds…
“quiet” has been ruined by LLMs, so has “isn’t X, it’s Y”
But anthropic and the others don’t give subscriptions to enterprises, only tokens at full API prices
Not quite true, for low seat organizations there’s a fixed rate Teams plan
Yeah, but Accenture is anything but a low seat org.
I have a friend that works there. They told me about this, and I was like “per day?”
I was not prepared to learn that the budget was less than a dollar per day. Might as well just stop using AI agents that that point.
Might as well just stop using AI agents that that point.
Wouldn’t that be terrible…
For Accenture? Yes. They are professional bullshit shovellers. AI was a godsend for them.
Might as well just stop using AI agents that that point.
I don’t work at Accenture, but at a group that had a similar journey from adoption to scaling back to only the included tokens and I would say that is precisely the idea: giving up on ‘Agentic’ use of AI, but still have enough for some acceleration.
From what I gather, in our group the findings were:
- Obviously, it was costing a heavy chunk of change and then even more expensive when rates were hiked
- There was no good way for them to even know what productive end the spend went toward, they know which employee expended a bunch of tokens, but no idea what they would have used it on
- Despite a crank of volume of stuff, they weren’t seeing any business value of the stuff. Requests still took about the same time to fulfill, etc
- The most eager AI producers got ahead of their skis, and their stuff was useless because nothing and no one could reliably fix/tweak it after a certain point.
- All the scary ‘AI will ruin IT security’ that the AI companies themselves have been expressing likely didn’t help either.
- Perhaps the most uncomfartable of all, they found out that however good or bad an AI project turns out to be, management can’t tell the diference until it is too late. Being able to ‘manage’ the content has always been flawed, but the AI dials it up to eleven by ‘looking’ just right as a demo to someone that understands neither the customer nor the process of creating the work, but blows up in their faces in delivery. Kind of like how so many technology article writers breathlessly talked about the photoshop clone, despite many of them never having touched the relevant software, then heavy users reporting that it is actually shit, but the people writing about it have no context to know, but felt compelled to talk about it anyway. Those are management in the fields using AI.
Ooh, it’s the : build a prototype story!
Dev, ok here’s the 2 week built prototype, slammed together wildly fast, no architecture, no seatbelts. Don’t push it in a demo.
Manager: this Looks like a finalized project! Except a) and b), fix that, you got a day.
Dev: we should productize it, to make it a solid product, it takes 3 months, then I can add a and b without it starting to smoke when used.
Manager: No, I already shipped it.
I just want to put “$29 per month” into perspective:
Imagine a white collar employee earning $100K per year. $100K / 250 working days is $400 per day. Payroll tax and benefits can easily add +50% to that, so it’s not a stretch to guess that our hypothetical employee costs the employer $600 per day.
$29 per 20 working days is about 1/4 of 1% of that cost.
Yeah, that’s like the breakroom coffee budget for the employee that month…
Yeah it’s basically useless for agentic work. Especially with frontier models.
I didn’t realize the business plans were token metered differently from the Pro plan I’ve been on for a while. $20/month, and it has session/weekly limits, but I rarely hit them, and I use it, a lot.
I do wonder if this will encourage businesses to branch out into running local models for simpler tasks.
You can get plans as a business too, but then it’s per user, so if your plan has $200 worth of “usage” per user and one uses $0, that does not mean another can use $400 and you still have to pay for the non-user’s $200, or you can use API pricing to pay for exactly what your users use, which for OpenAI and Anthropic is egregious. You get like 10x more per dollar with the Chinese models. And some of those providers ALSO have plans that make it much cheaper.
Nevrr forget, Accenture used to be Arthur Anderson, of Enron infamy. They deserve any financial ruin they get, IMO.
Sheeit thanks for the reminder (and trip down memory lane)
Fuck AA
In my company some of the top users spend over $3000/m. $29 is maybe 30 mins - 1 hour continuous running for a *medium" sized model like Sonnet. Not sure about cheaper models.
I spent $200 in one session the other day, watching the clanker design, implement, and review the change. Still have no idea what I did but my name is on the commit
I’ve no idea how to even see that. There’s a little bar on the settings page on the Claude site that shows a percentage of usage, but I’ve genuinely no idea how much it’s costing - and since I’m looking for an exit strategy I also don’t really care. My company made it abundantly clear that they wished to fuck me non-consensually in the arse back in May and I’m done with it at this point.
Past month I’ve been “working” on a project I have zero care for, just burning Claude credits chewing through one stupid idea after another and signing off on it. No one is asking me to show anything for it, no one is checking in on anything. There’s a general apathy about everything, covered up with a typical corporate enthusiasm.
It’s all so fake and disgusting, lmao.
I have a menu bar plugin that shows it, don’t remember where I got it?
Sounds fulfilling.
/s
I hate it of course. I need to find an exit strategy
Don’t pay the bill. They’ll force you to quit before long.
Fair point, but not any less fulfilling than my day job already was, arguably maybe a little more fun to be spending the company’s money on nonsense. My job pays my bills very adequately, that’s normally the only fulfilling thing about it.
-
Interesting. Yeah, me and my coworkers spend about $1000 per month—so far. The bill keeps going up. One AI workshop thing they made us do spent $100 just in 1 session.
That’s basically saying that they shouldn’t use Claude at all.
Good.
