top 98 comments

sorted by: hot top controversial new old
[–] 137 points 1 month ago (3 children)

They’re working on it. They just need a little more of your internal documents and source code.

  • source
  • hideshow 3 child comments
  • [–] 61 points 1 month ago* (26 children)

    I'd wager it's simply not possible to be profitable using AI for most use-cases. You either burn way too many tokens on your queries or risk your business falling victim to an under-tokenized AI hallucination.

  • source
  • hideshow 26 child comments
  • [–] 26 points 1 month ago (18 children)
  • [–] 46 points 1 month ago (17 children)

    I love the assumption that if you spend enough tokens, it won’t hallucinate, like it isn’t a mathematical certainty or anything.

  • source
  • parent
  • hideshow 17 child comments
  • [–] 5 points 1 month ago

    I agree with you, but I liken to to how companies tend to replace competence with bureaucracy. With enough layers, a significant amount of mistakes get caught. It just slows everything down and costs a bunch of money. There are still problems that get through, just less of them.

  • source
  • parent
  • [+] [S] -12 points 1 month ago (14 children)

    You can cut hallucinations by 90% just by prompting the LLM to.not guess API and function names and to provide it with full docs or MCP to find docs.

  • source
  • parent
  • hideshow 14 child comments
  • [–] 19 points 1 month ago (12 children)

    Yeah but it’s not guessing in the first place. It has zero reasoning skills. It just outputs the most likely next word. Even when you ‘tell it not to’

  • source
  • parent
  • hideshow 12 child comments
  • [+] [S] -9 points 1 month ago (11 children)

    I know I am getting down voted for this because people haven't tried it. And let me differ, it has reasoning skills, it now gets beyond predicting the next word. And even reducing it to next word is some kind of primitive reasoning.

    By experience, by telling the AI to not guess APIs and function names and instead provide it with a way of looking through docs, and inspecting the code and other tools to get it do some tests, the hallucinations disappear. I've been able to have better results with this setup.

    This is very different than the "no mistakes" joke.

  • source
  • parent
  • hideshow 11 child comments
  • [–] 12 points 1 month ago* (last edited 1 month ago)

    "it has reasoning skills" Nope. It has larger context windows and someone coded in the ability to reprompt itself for "thinking" steps that are literally just a new prompt with the same context that gets fed back in which merely increases the chances of it guessing the next output token correctly.

    It still doesn't "reason". It still doesn't know what the words mean. The modified algorithms are just attempting to generate the corrective prompts for you. If you've ever messed with them long enough, you'd understand that these, "no wait fix it" prompts have rapidly diminishing returns at the best of times. Their "thinking" is a bandaid, and a piss poor one at that.

  • source
  • parent
  • [–] 9 points 1 month ago (2 children)

    I know I am getting down voted for this because people haven't tried it.

    People here have tried it, LLM use is basically enforced at a ton of companies and the tech industry crowd is a huge portion of the population here.

    You're likely being downvoted as you seem to have a fundamental misunderstanding of how LLMs function. LLMs don't reason. LLMs don't think. Telling an LLM to not guess might help ever so slightly, but LLMs can and do hallucate search results regularly.

    It's a mistake to trust that LLMs aren't hallucinating just because you told the LLM to not guess.

  • source
  • parent
  • hideshow 2 child comments
  • [–] [S] 0 points 1 month ago (1 child)

    I juste don't want to talk about this any more becausefor me adding those instructions in AGENTS.MD and giving it the full API and language and platform docs dropped significantly the hallucinations. You just disregarding my experience and want me to believe in your assumptions.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 5 points 1 month ago

    It's not an assumption. You can give an LLM all of the instructions in the world and it can still hallucinate. This is the fundamental flaw with LLMs, they're always hallucinating. They just have guidelines which make it so they get things right sometimes. They are not to be trusted just because you tell it to search or use agents.md.

    You still have this fundamental misunderstanding. I encourage you to actually look into how LLMs work. Do not use an LLM to do this research for you.

  • source
  • parent
  • [–] 2 points 1 month ago

    I’ll back you up on this. I’ve had the experience of seeing Claude fabricate stuff, and then when I challenge it, it says something like “I guessed, let me look deeper” and then it indeed does check deeper with docs or a configured LSP etc and then discovers the real answer.

    It doesn’t matter if this was the result of reasoning or just more pattern prediction, the result is that if you tell it not to guess, it responds appropriately.

  • source
  • parent
  • [–] 3 points 1 month ago (5 children)

    It helps, sure. But it still does stupid shit CONSTANTLY.

  • source
  • parent
  • hideshow 5 child comments
  • [–] [S] -3 points 1 month ago (3 children)

    It doing sgupid stuff is another issue separate from hallucinations. Preventing LLM from doing dumb shit is more difficult, you need more guards and harnesses.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 6 points 1 month ago (2 children)

    I don’t agree it’s a separate issue. Sure you can add in more guardrails, and spend more time making the best prompt. But no matter what you do, at some point it WILL do something you either didn’t ask for, or explicitly told it not to do.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 6 points 1 month ago (6 children)

    This is kinda fundamental to economics: the moment AI learns to perform some task, that task becomes less valuable.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 1 point 1 month ago (5 children)

    But then aren't there just different tasks to do? VisiCalc and Excel didn't make accountants irrelevant.

  • source
  • parent
  • hideshow 5 child comments
  • [–] 4 points 1 month ago (4 children)

    That's the economists angle, yes, but that's not what I'm getting at.

    You can't make a business with AI because anyone can use an LLM to do anything an LLM can do.

    There are exceptions in the short term, but they may not last. For example, we're paying a hefty subscription fee for an AI service that scrapes financial data from scanned pages. It's much more accurate than traditional OCR.

    Presently its better than its competitors so commands a premium price. However, I don't think it will enjoy that position for long.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 1 point 1 month ago (3 children)

    This doesn't make a lot of sense. Just because everyone can use Word doesn't mean you can't make a business using it. Oh no, everyone has calculators, guess you can't have a business that uses calculators. Businesses and people in general use the best tools available for the job, which might be pencil and paper, or eventually an LLM.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 2 points 1 month ago (2 children)

    You're conflating a tool with a service.

    If you noticed that calculators were really good at adding up numbers, and started a business using a calculator to add up numbers for people, you won't attract any customers because people can just use their own calculators.

    That does not mean businesses do not use calculators.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 1 month ago (1 child)

    Wild, accountants must not exist in your mind. Everyone can own a computer but there are massive services of renting those computers 'in the cloud'. Everyone can buy storage, but for some reason the majority think its easier and cheaper to buy in the cloud. Hell, you could go to the store and purchase a movie, but many still think seeing them in a theatre, or you know, streaming them are a better option. You say LLMs have no business opportunity which seems to show the failure of your thinking.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 58 points 1 month ago (2 children)

    In a gold rush it is better, on average, to sell the shovels then to mine the gold.

  • source
  • hideshow 2 child comments
  • [–] 120 points 1 month ago (7 children)

    Because it is a scam. Just like the people selling, instead of using:

    • healing crystals
    • essential oils
    • blockchain
    • penis-enlargement pills
    • JIRA
  • source
  • hideshow 7 child comments
  • [–] 14 points 1 month ago (3 children)
  • [–] 7 points 1 month ago (1 child)
  • [–] 8 points 1 month ago

    I mean yes, I hate the manager who set up JIRA that way, but JIRA is marketed to and designed for managers to over-complicate it. It is not designed as a tool for engineers to track their work. It is designed for managers to mis-manage teams, because that sells at volume.

    I also have other reasons for hating JIRA:

    1. no ticket hierarchy
    2. no multiple owners
    3. slow. so very slow
    4. sometimes keyboard shortcuts work, sometimes they don't. There is no apparently pattern.
  • source
  • parent
  • [–] 32 points 1 month ago* (2 children)

    They are going for the gold rush pick axe business model

  • source
  • hideshow 2 child comments
  • [–] 31 points 1 month ago (2 children)
  • [–] 25 points 1 month ago

    Been shouting this from the rooftops since the beginning. If it was half as good as claimed, I would NOT have access to it.

  • source
  • [–] 17 points 1 month ago (2 children)

    If you come to seminar at the Holiday Inn, I'll teach you why! Only $500, and at the end you'll be able to start your own AI powered business and make millions per year!

  • source
  • hideshow 2 child comments
  • [–] 34 points 1 month ago* (1 child)

    If I recall correctly, shovel companies made a lot of money while only a few of their shovels actually struck gold.

  • source
  • hideshow 1 child comment
  • [–] 18 points 1 month ago

    Recent scholarship confirms that merchants made far more money than miners during the gold rush.
    The wealthiest man in California during the early years of the rush was Samuel Brannan, a tireless self-promoter, shopkeeper and newspaper publisher. Brannan opened the first supply stores in Sacramento, Coloma, and other spots in the goldfields. Just as the rush began, he purchased all the prospecting supplies available in San Francisco and resold them at a substantial profit.

    https://en.wikipedia.org/wiki/California_gold_rush

  • source
  • parent
  • [–] 15 points 1 month ago*

    I saw a thread a few minutes ago on a profitable AI business. The business model was hacking websites at scale to steal hundreds of thousands of credit card numbers...

    https://lemmy.world/post/52304304

  • source
  • [–] 30 points 1 month ago (5 children)

    Better question is if you can really build an app as easily as writing a prompt, why would anyone pay you for your app?

    Somehow AI bros like to pretend that AI is some secret lifehack despite it being one of the most mainstream things in human history.

  • source
  • hideshow 5 child comments
  • [–] 1 point 1 month ago (4 children)

    It still takes time and effort to be a project manager for app developer agents. Might as well just pay for it and get stuff done today. Instead of spending time on software maintenance, you could just pay for someone else to deal with all that hassle.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 2 points 1 month ago (3 children)

    Right. But Fermi's Paradox applies.

    Everyone is buying and selling software miracle cure...but all of the software itself is still exaclty as healthy as a con artists prize race horse (not healthy, and actually a mule).

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 1 month ago (2 children)

    Fermi paradox? The one about aliens existing?

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 1 month ago (1 child)

    Yea. Slightly modified - "If conditions are now so prefect now that 8 billion humans can ask their computer nicely to program itself, and it works, then why is non-shit-tier software still vanishingly rare?"

  • source
  • parent
  • hideshow 1 child comment
  • [–] 2 points 1 month ago

    Oh that way. Now I get it.

    Anyone can prompt Shlawp to write some code, but an expert would need to check it for errors and stupid design decisions. Normally, the user isn’t qualified to verify the validity of the answer or quality of the code and that’s why people who use Shlawp can’t make high quality applications.

    So, when you say “it works” it’s true for being able to generate at least some kind of code, not good code. The thing about Shlawp is that all of its outputs are very mediocre at best. They look convincing to everyone who aren’t qualified to evaluate the outputs. Those who are, will immediately notice all sorts of flaws.

  • source
  • parent
  • [–] 37 points 1 month ago (7 children)

    When potatoes came over from the New World, a lot of Europeans grew them as decorative plants. New shit makes people do weird stuff.

    Not everybody is trying to make a business with so-called AI. The peddlers are just the loudest.

    The companies promise their models can do all this stuff for paying customers. If the models were truly that capable, the so-called AI companies ought to be doing everything with them and be raking in the dough like mad. Which they're not because they know the limitations of the models. It's what economists call the bullshit gap between marketing and reality. And by economists I mean me.

  • source
  • hideshow 7 child comments
  • [–] -2 points 1 month ago (5 children)

    I mean if you were an economist you'd know exactly why AI companies are making massive profits, because its more profitable to sell shovels during the gold rush than it is to find gold. Doesn't mean gold is worthless and not profitable, it's pretty basic economics, but you know that already.

  • source
  • parent
  • hideshow 5 child comments
  • [–] 8 points 1 month ago (3 children)

    I mean if you were an economist you’d know exactly why AI companies are making massive profits

    They are literally not making massive profits. They may be making lyin' Sam Altman rich, but the VC money is drying up and OpenAI has repayments coming up in 2027 that it doesn't have a chance in hell of being able to payback.

    https://www.wheresyoured.at/brokenomics/

  • source
  • parent
  • hideshow 3 child comments
  • [–] 23 points 1 month ago (2 children)

    This is exactly the question i ask "investment advisor" services when they call asking if i want help managing investments.

    For you to have a system that works well enough for me to be interested , there'd be no need for you to spend money on call centers full of people ringing up strangers for money, you could take all of that money and invest it in the plan you are trying to sell me.

  • source
  • hideshow 2 child comments
  • [–] 3 points 1 month ago (1 child)

    I've always thought that as well. Surely it wouldn't take them long to become ludicrously wealthy with their top investing acumen.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 3 points 1 month ago

    Ikr. If I knew a "foolproof" investing strategy to "make loads of money", I wouldn't spend my time trying to teach other people about it, id just be using that strategy myself and actually NOT want other people to know about it.

  • source
  • parent
  • [–] 11 points 1 month ago (13 children)
    1. The AI tech they peddle is impressive and has some uses; but it is nowhere near ready to live up to the hype they created in which you can just pay $10K a year for a license and get rid of a $250K employee

    2. The AI tech they sell for $10K a year, actually costs $20K to run and they know nobody would pay the full cost, let alone the cost + their profit margin

    PS: those are made up numbers to answer the question, while not real numbers, they represent the points I am trying to make

  • source
  • hideshow 13 child comments
  • [–] 15 points 1 month ago

    That’s because the whole point of AI is to finally back us all into a corner where cloud computing is our only option and we can be monitored and controlled, our news and opinions filtered, dissenters found quickly, and our IP hoarded and summarized, our markets predicted and our products beaten before they launch.

    They dazzle the brain dead middle managers who all lament the lack of flying car futures with a piece of magic, and once enterprise (the big money) isn’t demanding powerful workstations anymore then average consumers can get fucked and we will be locked out of society if we organize.

  • source
  • parent
  • [–] 2 points 1 month ago (10 children)

    The AI tech they sell for $10K a year, actually costs $20K to run

    Open AI and anthropic make a profit on inference / api usage. It's only really the training that's dragging down their bottom line.

  • source
  • parent
  • hideshow 10 child comments
  • [–] 6 points 1 month ago (9 children)

    Well unless they have reach a level where training is not needed (which will be never for an LLM), what you told me is that you make a profit freezing ice cream but lose it on the eggs, cream and sugar necessary to make it

  • source
  • parent
  • hideshow 9 child comments
  • [–] 2 points 1 month ago

    The model releases are getting shorter and shorter. close to a month now. There is no actual profitability not because the price isn't higher than cost of ingredients, but because they don't count throwing away the ice cream mixer, and factory every month, when they should if replacing them is part of the process.

  • source
  • parent
  • [–] 1 point 1 month ago* (last edited 1 month ago) (7 children)

    Training is not needed to run inference on the existing models, it's needed to make new models. If LLMs plateau and there's no use in training new models, or the government regulates them and they can't train new frontier models, then they can still run inference on the currently existing models and make a profit.

    This is more like making a profit for making ice cream but losing a bunch of money researching new recipes for ice cream. If you decide the recipe is good enough and stop doing that research you can still make money making and selling ice cream.

  • source
  • parent
  • hideshow 7 child comments
  • [–] 0 points 1 month ago (6 children)

    yeah? so a model trained in 2 year old data is just as good? as a model trained today?

    If you decide the recipe is good enough and stop doing that research you can still make money making and selling ice cream.

    So they are just choosing to lose money? they have a perfectly good, profitable product but they choose to lose money... ok bud

    Your point only makes sense if we agree the current models are still not good enough to make a profitable product and thus, they need to keep pushing

  • source
  • parent
  • hideshow 6 child comments
  • [–] 0 points 1 month ago* (last edited 1 month ago) (5 children)

    so a model trained in 2 year old data is just as good? as a model trained today?

    Yes, assuming a plateau in LLM capability the only reason you'd train on an updated corpus is to get fresh info. That's not worth it because:

    1. The model will most likely be wrapped in a harness that has search capabilities, so it can use that to fetch fresh info
    2. 2 year old data may actually be better as the Internet becomes increasingly tainted with LLM content

    Most of the improvement from new models isn't coming from expanding or updating the corpus, its coming from increasing the parameter size and reinforcement learning.

    So they are just choosing to lose money?

    No, they are competing. If open AI decides to stop training new models right now then anthropic will and take all there business as the switching cost for models is low. That's why they want the government to regulate it, so they can have a ceasefire to start taking in profits from inference without worrying about their competitor making a new better model and eating there lunch.

  • source
  • parent
  • hideshow 5 child comments
  • [–] 1 point 1 month ago (4 children)

    if that were true, and the current models were actually good, there would be wide adoption of the current models while people wait for the next leap… instead, you get glimpses of the terrible numbers like 3% of user base is actually paying for these things

    maybe we torture our analogies to death but you make it sound like they have the formula for coca cola but instead of selling that, they are burning cash trying to find an even better formula

    when the reality is that they have something that is technically drinkable and not officially poison and they are selling it like the next elixir of life (while they burn cash trying to catch up to their own hype)

  • source
  • parent
  • hideshow 4 child comments
  • [–] 3 points 1 month ago* (last edited 1 month ago) (3 children)

    People are adopting the current models, mostly coders right now. Most people using LLMs are using them for basic information search, ie. Google AI overviews. Those tasks don't require the top models and don't burn that many tokens so the companies keep them free to get the public exposed to AI.

    Then there's the 3% of people who are the power users using it for work, especially coders. For productivity the top models do perform better and the stakes are higher so they need to perform better. The top models aren't needed for every task, but they shine as an orchestrator of smaller models handling more basic tasks. Coding also requires a lot more tokens, to read all the existing code; to do "thinking" which generates a bunch of output tokens to imitate reasoning; then to generate the code itself; then to review it, adjust after a review...

    All of that equates to large bills for token spend to anthropic or open AI. My Claude bill on my company account is approaching $2,000 this month, and that's about average / what's expected from an engineer at my company. I'd bet that every engineer in silicon valley is burning through a similar amount as well.

    This is why both open AI and anthropic are both seeing extremely high revenue growth. Anthropic ended 2025 with $9b in revenue , they are now on track to hit $100b revenue this year, for reference / the analogy coca cola made $47b this year, it's a larger revenue then every other software company except Microsoft. THAT IS INSANE, no company has ever seen that kind of growth. Yes it's being weighed down by training costs so they aren't profitable but if they even double there revenue next year that could change.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 1 month ago (2 children)

    People are adopting the current models, mostly coders right now.

    Models are being push down people's throats, mostly to coders right now... there FIFY

    Most people using LLMs are using them for basic information search, ie. Google AI overviews. Those tasks don’t require the top models and don’t burn that many tokens so the companies keep them free to get the public exposed to AI.

    Again, this is forced, you literally cannot perform a google search (or bing) without getting some AI results whether you want it or not. Then AI companies, like you, use this to pretend it's actual demand for their products

    ... All of that equates to large bills for token spend to anthropic or open AI

    If 3% of people generate the cost the AI companies cannot keep to the prices they charge, then this is the definition of an unprofitable business

    This is why both open AI and anthropic are both seeing extremely high revenue growth. Anthropic ended 2025 with $9b in revenue , they are now on track to hit $100b revenue this year

    No, this is what they ARE CLAIMING THEY WILL REACH. Last year, Altman said it was ridiculous that people were saying they generated $12 billion in revenue... just to then show it was less than that. Now they are delaying their IPO which means the books stay closed for another while longer

    THAT IS INSANE, no software company has ever seen that kind of growth

    Again you are just drinking the koolaid here (not counting all the anecdotal evidence you now apply to all of Silicon Valley). The truth is that whatever revenue they actually have (or claim, "trust me bro"), you can't even know if half of it comes from new investments, creative accounting or selling off assets, you just assume it is more people demanding more of their services and paying through the nose for it even if they have barely showed any tangible results in real life https://fortune.com/2026/08/22/executives-ai-productivity-layoffs-study/

    but if they even double there revenue next year that could change.

    Oh that's it? just double the already inflated revenue they are reporting without any evidence? of course!!! and all I need to fly is to grow some wings, that's it, it's really that simple

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 1 month ago (1 child)

    I don't think you understand the scale, even if half of that number is real that's $50b , that's larger then coca cola, that's larger then Salesforce and every other software company except Microsoft. That's 5x growth on $10b , that is unheard of in the software industry, much less any other industry.

    IDK what to tell you if you don't believe the numbers, yes they aren't official but every credible source is pointing to massive revenue growth. We'll see the actual numbers in the coming weeks as they prepare for the IPO, and if they do turn out to be cooking the books to the tune of $50b , I'll come back to this comment and admit I'm a credulous idiot who fell for the hype. But if they are true I'll also be coming back with an I told you so.

    Also do you think they would cook the books to that extent if they are still going ahead with an IPO? Fabricating $50b in revenue would hurt an IPO once it's found out way more then some temporary hype from a NYT article.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 1 month ago

    that's larger then coca cola, that's larger then Salesforce and every other software company except Microsoft. That's 5x growth on $10b , that is unheard of in the software industry, much less any other industry.

    and you only need the to double it (the full 100 billions} to make a profit lol…

    IDK what to tell you if you don't believe the numbers, yes they aren't official but every credible source is pointing to massive revenue growth

    define "credible"

    Also do you think they would cook the books to that extent if they are still going ahead with an IPO? Fabricating $50b in revenue would hurt an IPO once it's found out way more then some temporary hype from a NYT article.

    the ipo they are already postponing for "safety"?

    https://fortune.com/2026/09/12/sam-altman-openai-ipo-delay-ill-advised-moment-safety-concerns/

    https://www.investors.com/news/technology/anthropic-ipo-delayed-openai-expects-massive-cash-burn/

  • source
  • parent
  • [–] 9 points 1 month ago

    They're waiting for the to big to fail bailout like the banks in '08.

  • source
  • [–] 8 points 1 month ago

    AI is becoming a pretty big problem where executives are challenging experts or subject matter experts on topics based on their Ai usage. The problem is because the answers sound really good, but often the prompts aren't quite right so then the experts need to do a bunch more work to explain and justify things that the executives just find rebuttals for through further ai prompting.

  • source
  • [–] 8 points 1 month ago

    For the same reason that people were selling crypto mining boxes rather than just plugging them in:

    The only guaranteed profit in a gold rush is to be the one selling shovels.

    See Nvidia for details.

  • source
  • [–] 7 points 1 month ago

    You're on to something …

  • source
  • [–] 6 points 1 month ago

    Because you need to have idea that people agree to pay money for. AI companies have idea that is selling GPU (nvidia) computer generated chat ( OpenAI, Anthropics ). People don't really need GPU or computer chat but they are willing to pay for it because they believe it will help them to save time or money by spending time and money using this shit.

    That's paradox of modern times. We are being sold time spending tools that bring nothing to our lives except turning our heads from the fact that planet is dying.

  • source
  • [–] 5 points 1 month ago

    We all know why…

  • source
  • [–] 5 points 1 month ago* (3 children)

    AI is highly subsidized for users so AI wrappers can be profitable. (Even though they are a net negative to the economy)

    But ai companies burn trillions of dollars of investor money with fake promises. They can't make a reddit chatgpt bot that charges a user 10$ a month and make a profit.

  • source
  • hideshow 3 child comments
  • [–] 1 point 1 month ago

    Google seems to be figuring it out pretty quickly. Getting easier and better to simply use Gemini than it is to actually use Google's search. That's been the tech idea for years. Come in quick and disruptive, make it cheap on someone else's back, and then when people are hooked, wrack up the costs.

  • source
  • parent
  • [–] 9 points 1 month ago

    Cuz it doesn't work.

  • source
  • They want somebody else to be on the hook for their screw-ups. Doesn't mean they won't get sued if it's something bad enough or their meatshield business decides to make an issue out of it but this way, they've reduced their risks a little.

  • source
  • load more comments
    view more: next ›