Yeah, I watched Veep either during or after the first Trump presidency. The situations that were supposed to feel absurd felt borderline sane in comparison to reality.
oh shit, I thought it's per year initially, I am considered a senior dev in a lot of setups (8yoe), worked remotely for European companies (hence European standards on code quality), hold PhD yet my monthly salary didn't cross 4k yet, you can hire an army of me with this.
Yeah I was about to say. I'm an American software engineer at 10 years in the industry. That number is roughly my yearly salary, and they're talking monthly, so that's like 12 of me. From what I've noticed in online job searches, European positions of roughly equal experience and skill levels get paid somewhere between 30-50% less. Is that because of the stronger welfare systems? Or maybe the differences in cost of living, which could vary based on country and region?
The cost of living varies a lot not only from country to country, but from city to city within the same country. But yes, it's the safety net that we have here. You don't go bankrupt if you have a medical emergency, your health care insurance isn't tied to your job, child support is better (varies by country), worker rights are stronger (varies by country, but on average it's better than the US), secured paid maternity leave, vacations, sick leave, etc. Generally speaking, if you live in Europe but work for a US company, you basically hit a lottery. As a software dev, at least.
FWIW a rule of thumb is that an employee costs a (US) company roughly twice what their annual salary is because of benefits, administrative costs, office space etc. That's still like 6 human engineers.
I worked for a chinese startup in Beijing and we barely employed git or reviews, massive projects had 0 tests, we put 0 thoughts on architecture, just stop when it works.
I joined an established European company after that and the difference was huge, I had to relearn everything, proper code reviews, follow best practices, you don't stop when it works, you stop when it works, it's well tested (extensive business tests plus at least one required integration test for every change you make), iterated over the architecture a few times to have something simple to understand and clean.
I was also a contractor for another very new European startup where the process was not as robust but still better than the first one.
Congratulations, you make around as much as a barista in Boston! Wooo
There are developers making over 100k a month just building AI at these AI companies. The entire point is to eliminate anybody doing knowledge work that isn’t at the bleeding edge of understanding. It’s commoditizing a skill that commanded a huge premium, but is way overstated in HCOL areas.
If it isn’t AI taking the tech jobs, it’ll be India and other countries where they make even less than you but can do the job because they are no less capable. We’re all just meant to be servants to our billionaire masters anyway. They don’t care how many people have jobs, so long as that they pay as few of us as little as possible to get what they want. They don’t care if a handful of us are paid like royalty so long as it’s less than how much they would have paid all of us otherwise.
That's insane. A company I work at leverages AI heavily across sales, product, engineering, etc.
Company of 50ish.
Nobody I know at our size or smaller uses it as heavily as us. I realize that's subjective.
We're sub $20k/mo.
This person is not using a tool correctly. They're probably burning on inference. Probably inflated rules and context. Probably shipping slop and then fixing slop with slop. Absolute garbage.
Their company is providing AI tooling, and the clients are paying for it from what I can see. He's "proud of the invoice" because his company is earning a margin on effectively reselling it. The more they paid Anthropic the more their profit was.
My company is going wild like this too. I mean, it makes sense - more AI use means code is being written faster and thus we’re making more money.
But they go weirdly quiet when devs ask if AI has actually been speeding up the time to get a finished product in front of customers. You know, the only metric you should care about?
That's an... interesting correlation they're making, more code = more money. I know it's not you personally making that comparison, but man is it strange. That's a very business school way of thinking.
What good is "more code" from the LLMs, if I have to scrutinize it for bugs and vulnerabilities? More code only means more surface area, more points of failure. And of the AI I've tried, every single one writes far far far too much code. And all that time in code review, QA, user acceptance testing, that absolutely does not make the company more money - it costs them more money, in paying for labor. And it doesn't get the product to the end user faster anyway.
I'm just ranting and this a minor point, but speed is also not the only metric I would care about. I'd also care about making sure the user doesn't experience many bugs - preferably no bugs at all. The classic engineer's triangle still holds: "Fast, Cheap, and Good: choose 2." And AI seems to pick "Fast" twice. XD
More code only means more surface area, more points of failure. And of the AI I’ve tried, every single one writes far far far too much code. And all that time in code review, QA, user acceptance testing, that absolutely does not make the company more money - it costs them more money, in paying for labor. And it doesn’t get the product to the end user faster anyway.
Duh, just have the LLM do code review, QA, and testing for you! And then blindly ship it to production once that's done.
I mean, it makes sense - more AI use means code is being written faster
This AI you speak of - is it in the room with us right now?
Because if you mean to say LLM slop machines, then "code is being written faster" does not imply the code is in any way useful.
I use an LLM as a glorified search engine at work. Basically to search thousands of pages of documentation and tickets to get context before approving something. 2-3 prompts amounts to a dollar. Imagine doing that several times a day multiplied by thousands of employees (because everyone is mandated to use AI). And that’s a simple task, not even coding yet.
Nice. I've been making various models create QR code generators that only work within a terminal. Then they collapse when I tell them to make the code hardened against exploits.
It's a company that sells 'AI' services to their customers. Most likely the major part of the 113,000 dollars is paid by those customers. They're basically resellers.
The company I work for (publicly listed) has granted company-wide access to Copilot, but requires users to complete a pretty extensive (2+ hour) training module before you’re able to use it.
Pretty obvious stuff like, don’t upload sensitive data, validate output etc. I just use it as a glorified search engine usually - validating function syntaxes etc., and also cot rewriting email to be less blunt.
Really comes down to whether companies have; no tech departments to stop employees from using AI, small departments that don’t want to deal with the headache, large departments capable of managing the rollout.
I wonder how they feel about the fact that Microsoft declared Copilot to be "for entertainment purposes only", therefore shielding themselves from any lawsuits and leaving all users liable for the mistakes Copilot makes
Pretty unsurprised, I imagine - our training pretty much instructs us to not take anything it produces as gospel already.
Though given how much the company is getting charged for it, I think there may be some reconsideration in the budget for next financial year - given Microsoft have all but confirmed that it’s not much more than an overly verbose Magic 8 Ball
Given that the rest of our annual training refreshers are 15 - 30 minutes, and that I can boil down the training to 2-3 bullet points here - yes, 2+ hours is (relatively) extensive.
I'm a programmer analyst(don't want to say where). I just notice on most job postings that AI is always being looked at as a tool for someone to use so i was just curious.
Yeah, we created a new buzzword that means the same thing as "autonomous".
Okay, well, it's actually less good than autonomous, because you have to babysit the thing to make sure that it doesn't shit the bed. But totally worth it! Because it's agentic!
89 Comments
AbsolutelyNotAVelociraptor@piefed.social · 284 pts · 131d
"Here's why:"
...
Is it because you're dumb?
UnderpantsWeevil@lemmy.world · 134 pts · 131d
It's cause he's a bot account doing marketing for Anthropic
zurohki@aussie.zone · 47 pts · 131d
It's to "Swan AI", so I assumed they were happy about the invoice because they're reselling that service to suckers for a lot more than that.
UnderpantsWeevil@lemmy.world · 17 pts · 130d
Scamming as a Service
msage@programming.dev · 2 pts · 130d
Always has been.
zakobjoa@lemmy.world · 14 pts · 130d
That phrase alone should land you in prison.
ThunderComplex@lemmy.today · 8 pts · 130d
Sad hbomberguy noises
Passerby6497@lemmy.world · 3 pts · 130d
That one is almost as bad as the videos that start out with DiD yOu KnOw
a_non_monotonic_function@lemmy.world · 9 pts · 130d
With a four person team spending over a million annually (potentially) on AI tokens is morbidly stupid.
dev_null@lemmy.ml · 8 pts · 130d
They are reselling it, so it's their clients who are actually spending the money. Meanwhile their 4 person team is earning all the margin.
a_non_monotonic_function@lemmy.world · 1 pts · 130d
That sounds even more remarkably stupid.
The image also doesn't appear to provide any of that context.
Reverendender@sh.itjust.works · 226 pts · 131d
ech@lemmy.ca · 122 pts · 131d
I would love another season or two of Silicon Valley to roast the latest techbro bullshit.
Reverendender@sh.itjust.works · 57 pts · 131d
It was an incredibly prescient show, too
TrickDacy@lemmy.world · 18 pts · 131d
Mike judge always delivers.
SCmSTR@lemmy.blahaj.zone · 3 pts · 130d
He working on anything else these days?
TrickDacy@lemmy.world · 9 pts · 130d
https://en.wikipedia.org/wiki/Common_Side_Effects
He's executive produces that show and does some voice work on it. I think it's a really great show.
mycodesucks@lemmy.world · 3 pts · 130d
Beavis and Butthead and I assume the King of the Hill revival.
Rothe@piefed.social · 20 pts · 130d
I think it would suffer from reality being so batshit insane that it would struggle to surpass it in any comedic way, just like The Onion these days.
jacksilver@lemmy.world · 7 pts · 130d
Yeah, I watched Veep either during or after the first Trump presidency. The situations that were supposed to feel absurd felt borderline sane in comparison to reality.
ech@lemmy.ca · 3 pts · 130d
If anyone can make it work, it's Mike Judge.
Blackmist@feddit.uk · 4 pts · 130d
It feels like it already is.
criss_cross@lemmy.world · 166 pts · 131d
My company has been jizzing themselves that our token usage has gone up.
I’ve never seen so many people rushing to add a middleman into their product dev.
umbrella@lemmy.ml · 6 pts · 130d
slaacaa@lemmy.world · 122 pts · 131d
Tokenmaxxing
Ramenator@lemmy.world · 101 pts · 131d
You could hire twelve senior engineers in Europe for that kinda money
idriss@lemmy.ml · 55 pts · 131d
oh shit, I thought it's per year initially, I am considered a senior dev in a lot of setups (8yoe), worked remotely for European companies (hence European standards on code quality), hold PhD yet my monthly salary didn't cross 4k yet, you can hire an army of me with this.
Opisek@piefed.blahaj.zone · 29 pts · 131d
Oh yeah, but can an army of you draw me as a Ghibli character? Checkmate, liberals /s
Rooster326@programming.dev · 5 pts · 130d
Anthropic doesn't have an image generator so neither can they
Valmond@lemmy.dbzer0.com · 1 pts · 130d
I can, but I'll also cost way less if you hire me for that instead of coding.
HarkMahlberg@kbin.earth · 9 pts · 131d
Yeah I was about to say. I'm an American software engineer at 10 years in the industry. That number is roughly my yearly salary, and they're talking monthly, so that's like 12 of me. From what I've noticed in online job searches, European positions of roughly equal experience and skill levels get paid somewhere between 30-50% less. Is that because of the stronger welfare systems? Or maybe the differences in cost of living, which could vary based on country and region?
AbsolutelyClawless@piefed.social · 13 pts · 131d
The cost of living varies a lot not only from country to country, but from city to city within the same country. But yes, it's the safety net that we have here. You don't go bankrupt if you have a medical emergency, your health care insurance isn't tied to your job, child support is better (varies by country), worker rights are stronger (varies by country, but on average it's better than the US), secured paid maternity leave, vacations, sick leave, etc. Generally speaking, if you live in Europe but work for a US company, you basically hit a lottery. As a software dev, at least.
ChickenLadyLovesLife@lemmy.world · 9 pts · 130d
FWIW a rule of thumb is that an employee costs a (US) company roughly twice what their annual salary is because of benefits, administrative costs, office space etc. That's still like 6 human engineers.
Rooster326@programming.dev · 3 pts · 130d
Hello we would like to welcome you to the year of our Lord.
HarkMahlberg@kbin.earth · 1 pts · 130d
Ah right I knew that rule but forgot to apply it here.
ChickenLadyLovesLife@lemmy.world · 6 pts · 130d
Is that a thing? I worked for a mostly-Finnish company for a couple of years and their code base wasn't any better than anybody else's.
idriss@lemmy.ml · 9 pts · 130d
It could be just my biased view/experience.
I worked for a chinese startup in Beijing and we barely employed git or reviews, massive projects had 0 tests, we put 0 thoughts on architecture, just stop when it works.
I joined an established European company after that and the difference was huge, I had to relearn everything, proper code reviews, follow best practices, you don't stop when it works, you stop when it works, it's well tested (extensive business tests plus at least one required integration test for every change you make), iterated over the architecture a few times to have something simple to understand and clean.
I was also a contractor for another very new European startup where the process was not as robust but still better than the first one.
arcine@jlai.lu · 5 pts · 130d
In some countries in Europe, we're used to stating revenue monthly instead of yearly !
Prove_your_argument@piefed.social · 3 pts · 131d
Congratulations, you make around as much as a barista in Boston! Wooo
There are developers making over 100k a month just building AI at these AI companies. The entire point is to eliminate anybody doing knowledge work that isn’t at the bleeding edge of understanding. It’s commoditizing a skill that commanded a huge premium, but is way overstated in HCOL areas.
If it isn’t AI taking the tech jobs, it’ll be India and other countries where they make even less than you but can do the job because they are no less capable. We’re all just meant to be servants to our billionaire masters anyway. They don’t care how many people have jobs, so long as that they pay as few of us as little as possible to get what they want. They don’t care if a handful of us are paid like royalty so long as it’s less than how much they would have paid all of us otherwise.
AlecSadler@lemmy.dbzer0.com · 96 pts · 130d
That's insane. A company I work at leverages AI heavily across sales, product, engineering, etc.
Company of 50ish.
Nobody I know at our size or smaller uses it as heavily as us. I realize that's subjective.
We're sub $20k/mo.
This person is not using a tool correctly. They're probably burning on inference. Probably inflated rules and context. Probably shipping slop and then fixing slop with slop. Absolute garbage.
djmikeale@feddit.dk · 38 pts · 130d
From their website it looks like they're "reselling" so, that's my best guess as to why their bill would be so high
dev_null@lemmy.ml · 28 pts · 130d
Their company is providing AI tooling, and the clients are paying for it from what I can see. He's "proud of the invoice" because his company is earning a margin on effectively reselling it. The more they paid Anthropic the more their profit was.
SethTaylor@lemmy.world · 64 pts · 130d
Man, if the robots tried to take over this guy would really just drop to his knees and suck robot dick, wouldn't he?
Inucune@lemmy.world · 61 pts · 130d
Export the data, dissolve the LLC, leave the bill unpaid and start a new LLC using the data, repeat.
I don't think these AI companies have enough funding to take legal action without their part of the house of cards collapsing.
flop_leash_973@lemmy.world · 20 pts · 130d
This is the kind of snake eating its tail chaos I'm here for.
mal3oon@lemmy.world · 2 pts · 129d
Are there legal ramifications for doing so many times? It feels ... illegal? But fuck big tech either way.
jeffep@lemmy.world · 56 pts · 130d
In my school, kids who said "I'm proud of an invoice" got kicked in the balls and it was good for them in the long run.
i_stole_ur_taco@lemmy.ca · 45 pts · 131d
My company is going wild like this too. I mean, it makes sense - more AI use means code is being written faster and thus we’re making more money.
But they go weirdly quiet when devs ask if AI has actually been speeding up the time to get a finished product in front of customers. You know, the only metric you should care about?
HarkMahlberg@kbin.earth · 15 pts · 131d
That's an... interesting correlation they're making, more code = more money. I know it's not you personally making that comparison, but man is it strange. That's a very business school way of thinking.
What good is "more code" from the LLMs, if I have to scrutinize it for bugs and vulnerabilities? More code only means more surface area, more points of failure. And of the AI I've tried, every single one writes far far far too much code. And all that time in code review, QA, user acceptance testing, that absolutely does not make the company more money - it costs them more money, in paying for labor. And it doesn't get the product to the end user faster anyway.
I'm just ranting and this a minor point, but speed is also not the only metric I would care about. I'd also care about making sure the user doesn't experience many bugs - preferably no bugs at all. The classic engineer's triangle still holds: "Fast, Cheap, and Good: choose 2." And AI seems to pick "Fast" twice. XD
OwOarchist@pawb.social · 11 pts · 131d
Duh, just have the LLM do code review, QA, and testing for you! And then blindly ship it to production once that's done.
HarkMahlberg@kbin.earth · 2 pts · 130d
Silly me why didn't I think of that lol
raspberriesareyummy@lemmy.world · 10 pts · 131d
This AI you speak of - is it in the room with us right now? Because if you mean to say LLM slop machines, then "code is being written faster" does not imply the code is in any way useful.
Scubus@sh.itjust.works · 2 pts · 130d
/>thatsthepointtheyweremaking.jpg
bountygiver@lemmy.ml · 1 pts · 130d
Nothing has changed much huh? Plenty of companies uses lines written as a metric for productivity before all these.
Gerudo@lemmy.zip · 41 pts · 131d
But add another employee or 2, or God forbid increase benefits and they burn you at the stake.
fubarx@lemmy.world · 39 pts · 131d
Turret3857@infosec.pub · 34 pts · 131d
good business is when bankruptcy
scytale@piefed.zip · 28 pts · 131d
I use an LLM as a glorified search engine at work. Basically to search thousands of pages of documentation and tickets to get context before approving something. 2-3 prompts amounts to a dollar. Imagine doing that several times a day multiplied by thousands of employees (because everyone is mandated to use AI). And that’s a simple task, not even coding yet.
Mocheeze@lemmy.world · 9 pts · 131d
Nice. I've been making various models create QR code generators that only work within a terminal. Then they collapse when I tell them to make the code hardened against exploits.
9point6@lemmy.world · 26 pts · 131d
That's 12 fairly decent engineers' salary
Is AI really 4x-ing productivity at his company? Even the optimistic estimates don't even have it as a 2x improvement
It's baffling to me that people who's entire career hinges on them displaying a level of extreme competence, just tell on themselves like this.
kungfuratte@feddit.org · 19 pts · 131d
It's a company that sells 'AI' services to their customers. Most likely the major part of the 113,000 dollars is paid by those customers. They're basically resellers.
arcine@jlai.lu · 23 pts · 130d
You could have bought a small house with that kind of money 😅
DakRalter@thelemmy.club · 18 pts · 130d
In London, you might be able to buy a shed with that 😅
StupidBrotherInLaw@lemmy.world · 10 pts · 130d
You lot can afford sheds?!
Denalduh@lemmy.world · 11 pts · 130d
Look at this guy with shed money.
DakRalter@thelemmy.club · 2 pts · 130d
Heck no. I don't have that sort of money.
titanicx@lemmy.zip · 1 pts · 129d
I wish I lived where you do. Then again maybe not ...
arcine@jlai.lu · 1 pts · 129d
France, in a medium-sized city.
imjustmsk@lemmy.ml · 0 pts · 130d
SneakyWeasel@lemmy.ca · 20 pts · 131d
Well every job i have worked at requires me to not use ai. Is this all private companies?
thatKamGuy@sh.itjust.works · 24 pts · 131d
The company I work for (publicly listed) has granted company-wide access to Copilot, but requires users to complete a pretty extensive (2+ hour) training module before you’re able to use it.
Pretty obvious stuff like, don’t upload sensitive data, validate output etc. I just use it as a glorified search engine usually - validating function syntaxes etc., and also cot rewriting email to be less blunt.
Really comes down to whether companies have; no tech departments to stop employees from using AI, small departments that don’t want to deal with the headache, large departments capable of managing the rollout.
SethTaylor@lemmy.world · 7 pts · 130d
I wonder how they feel about the fact that Microsoft declared Copilot to be "for entertainment purposes only", therefore shielding themselves from any lawsuits and leaving all users liable for the mistakes Copilot makes
thatKamGuy@sh.itjust.works · 8 pts · 130d
Pretty unsurprised, I imagine - our training pretty much instructs us to not take anything it produces as gospel already.
Though given how much the company is getting charged for it, I think there may be some reconsideration in the budget for next financial year - given Microsoft have all but confirmed that it’s not much more than an overly verbose Magic 8 Ball
ChickenLadyLovesLife@lemmy.world · 3 pts · 130d
2+ hours is "pretty extensive" where you work?
thatKamGuy@sh.itjust.works · 10 pts · 130d
Given that the rest of our annual training refreshers are 15 - 30 minutes, and that I can boil down the training to 2-3 bullet points here - yes, 2+ hours is (relatively) extensive.
gwl@lemmy.blahaj.zone · 11 pts · 130d
Nah, you've just been lucky
What field you working in? I know some are more security conscious and don't allow AI for that reason (gambling, sports betting, banking)
SneakyWeasel@lemmy.ca · 4 pts · 130d
I'm a programmer analyst(don't want to say where). I just notice on most job postings that AI is always being looked at as a tool for someone to use so i was just curious.
Iheartcheese@lemmy.world · 20 pts · 131d
I really wanna see his reasoning
nimble@lemmy.blahaj.zone · 47 pts · 131d
Iheartcheese@lemmy.world · 40 pts · 131d
Fucking morons.
nimble@lemmy.blahaj.zone · 18 pts · 131d
9point6@lemmy.world · 15 pts · 131d
lol. lmao even.
HarkMahlberg@kbin.earth · 4 pts · 131d
As anyone who knows the value of compound interest can tell you, this is NOT a flex lmfao
zarkanian@sh.itjust.works · 3 pts · 130d
Yeah, we created a new buzzword that means the same thing as "autonomous".
Okay, well, it's actually less good than autonomous, because you have to babysit the thing to make sure that it doesn't shit the bed. But totally worth it! Because it's agentic!
ChickenLadyLovesLife@lemmy.world · 1 pts · 130d
God damn it, now I'm singing "agentic" in my head like "ozempic" in that fucking commercial.
jeffep@lemmy.world · 3 pts · 130d
I think I have bad news for you
rem26_art@fedia.io · 19 pts · 131d
this whole industry is completely unsustainable
motruck@lemmy.zip · 12 pts · 130d
Mandatory: CEOe are like the village idiot who won the lottery. Don't listen to them and stop giving them a voice.
Darcranium@lemmy.world · 4 pts · 130d
Yes this is so obviously copium
overkrill@lemmy.dbzer0.com · 9 pts · 131d
thats a lot of money for inherently Flaude Code
patatas@sh.itjust.works · 5 pts · 129d
This is 100% marketing for Anthropic. Think your monthly bill for the tokens you used is high? Rookie numbers
XaetaCore@lemmy.neondystopia.world · 5 pts · 126d
Never thought anyone could be so proud of sheer incompetence
ecards@sh.itjust.works · 1 pts · 124d
Money laundering?
GreenBeanMachine@lemmy.world · 1 pts · 130d