He was wrong about everything, the Claude S-1 proved him wrong.
>>110215218He is right.
>>110215227Wrong.
>>110215242
>>110215218Is he gonna commit suicide when Anthropic pulls $25 billion in Q4 with 85% inference margins?
>>110215242He’s just another luddite grifter
>>110215276>50% will fail by 2030that's a 50% underestimation
>>110215333Maybe
>>110215242is that from the same document where they censored all their debts?
>>110215218Reminder that this guy has ADHD
https://businesschief.com/articles/the-s-p-493-where-the-ai-booms-money-never-arrivedThe S&P 493, aka S&P 500 minus Mag 7 has seen ZERO productivity growth per employer, when you adjust for inflation.Your circular finance alchemy can only go so far.
>>110215242>revenue
>>110215242This got them boiling
>>110215218He will cope until the day of his death. I bet he will insist anthropic is committing fraud with false profit documents for the next 10 years. Anything but admit he is a grifter coper that's literally been wrong on everything. He's like a broken magic 8 ball that's stuck always saying "ai is gonna crash" and always being confidently wrong. The irony is that he is the kinda retard saying "you have to check ai output because it has le hecking hallucinations!", so rich.
>>110215343>luddite grifterThis. Sell your luddite stocks and buy AI stocks.
>>110215218he had his 15 minutes of fame with his overly contrarian stance. time to learn to nurse, Chud.
>>110217724Anthropic will not exist in 10 years.
>>110217945Two more weeks
>>110215333>Still has to subsidize tokens>Still not even remotely close to being profitable>Still has to borrow billions more for "growth"
>>110217949>Two more weeksuntil Anthropic is profitable
>>110217980Sad cope. Inference costs basically nothing.
>>110217970Subs are like 10pc of their revenue lol
>>110218067>data centers which require hundreds of megawatts of power>quickly depreciating, extremely expensive equipment>muh free inferencejej
>>110218261>conveniently avoids to mention the profits, and only talks about the costs >conveniently avoids to mention the salary of programmers, which will drop from $200k+ to $0
>>110218286>>conveniently avoids to mention the salary of programmers, which will drop from $200k+ to $0Just moving the costs somewhere else.
I like Ed. Seems like a nice smart guy. Anyone sharing their opinion near daily on the internet will get things wrong, forever to be archived, but he's one of the only people fighting against this insanity. I wish he would go deeper into the Anthropic EA and rationalist cults instead of going to hard into finance where he seems a little out of his depth. He still gets a lot right but no one is petty enough to compile a list of everything he got right like the petty fags who compile a list of the things he got wrong.
>>110215218The Anthropic S-1 proved him right
>>110218286Anthropic themselves said in their IPO that they have no path to profitability
>>110218319Incorrect, see>>110215242
>>110215218How convenient
>>110218333kwab
>>110218319>>110218324Is /g/ perpetually like this now? >>110217416E.g. Zero replies to this anon. The aim is clearly just to agitate. Any actual replies to posts like op or >>110215242 make zero effort to address logical points in replies and either say nothing relevant or just a simple insult disregarding the source. It's not even just AI topics. I'm not clear on why anyone thinks it's worth investing any time into shitting up /g/. What threat do they imagine needs to be quashed?
>>110218324Cope, Uber said the same in their IPO, it's just to cover their ass legally.
>>110218366This
>>110218319I'm not a businessfag. How normal is this for a S-1 document? Are they finalizing the numbers cause it seems wild not to list your obligations?
>>110218366>>110218390If a company knows it will be unprofitable and is legally forced to admit it's a possibility in its S-1, regardless of whether they believe it or not, that doesn't negate the fact it can or will be unprofitable. It just means they had to say it.
>>110218401If they could have disclosed them with confidence, they would have
I don't like AI and I think it is going to have incredibly negative consequences.But Zitron is still an obvious grifter and it's disappointing to see so many people taking him seriously or as some kind of authority. There are a lot of people who built an entire niche youtube career in the past 2 years on having a public existential crisis and arguing with themselves over AI, when really they have no interesting or special insights into anything.Zitron is just feeding a different category of people who just want to hear that it doesn't even work which is wildly more retarded, but comforting to a lot of people who want to believe they still have some value.
He said early 2024 that AI has peaked. He was already definitely proven wrong when o1-preview released September that year.https://www.wheresyoured.at/sam-altman-fried/
>>110215218the Claude S-1 proved him wrong.literally the opposite of truth lol?
>>110218440See>>110215242
>>110218429I've noticed most of the lists and examples people post proving Ed wrong are from 2024-2025. The fact I rarely see 2026 examples should be a lot more concerning for you, surely, for now shit is actually getting real. The tech back then was a toy for most people.
>>110218446I think it is more concerning that he is still given airtime as a serious analyst.
>>110218416Anon this is the equivalent of "THE SOFTWARE IS PROVIDED AS-IS" but in businessfag speak. It's the equivalent of "no warranty".
>>110218327incorrect see>>110218319
>>110218481He's been getting big platform airtime only in 2026, the year in which there are very few examples people post of him being wrong. That is WORSE for the bulls and it's why you're all squealing online about him. You hate seeing him appear on another big platform saying shit you struggle to disprove (now). In an ideal world for the bulls, he would've had a big platform during 2024-2025 where he could've made retarded claims on TV, but most of his shit back then was in his relatively small newsletter. The last thing you want is him on big podcasts saying (mostly) right things now, being a lot more careful with his wordings and research and getting a lot of acclaimed analysts and tech journalists working with, supporting and citing his work. He's starting to win and that's why AI bulls are panicking like fuck right now. If you genuinely thought Ed Zitron was a grifting retarded piece of shit, a lolcow of the AI era who'll make no impact, you would've see 10 viral tweets a day trying to take him down. You'd all let him continue to make a fool of himself. The fact you all feel the need to constantly drag him down so people don't take him seriously reveals an insecurity. I don't think any of you are as confident in AI as you pretend to be.
>>110218426This. I hate AI companies too, and I hate that my livelihood is in great danger because of this, but denying reality at this point is just a sad cope, and grifters like Zitron are selling copium to copers.
>>110218573I don't have any need to drag him down. I think this thread is a waste of time. His prominence is more an indictment of the leftist AI-psychosis that all of this is is actually fake and will come crashing down at any second.
>>110218573He's literally only right on a temporary basis. AI works. The technology could stop development tomorrow and it will still work as it stands now. The finances are irrelevant.
>>110218446He keeps making the same (incorrect) arguments. They proved him wrong in 2024-2025 when, as you say, ai was a toy. Now you could apply the same arguments when ai is already replacing programmers. Why don't you copers just admit reality at this point? This is just sad.
>>110218594Exactly this
>>110218591Ed on so many occasions has said the tech is good and has improved. He has nothing wrong with LLMs. He expects local AI to win long-term and has no problem with that for that's not destroying the lives of millions. You niggas enjoying the prices of tech right now? Make you feel good sucking Effective Altruist hairy cocks whilst you pay 5x the prices for inferior hardware? His arguments are laser focused on the industry on the whole. Where the big money is. He knows the technology is here to stay, he just fights for transparency in the industry, stronger financial regulation and for the valuations of all of these companies to be significantly reduced because it's just another version of cloud software, like cloud storage. He sees it as cloud intelligence basically. AI bulls constantly get confused about AI bears because they think the bears are trying to say the technology is a scam and doesn't work, when the bears are talking about the finance behind the technology. >>110218594Programming is the only industry heavily getting affected because it's programmers who are building the technology...for themselves. They know what programmers need out of LLMs and harnesses. They can test it right away in their own work and get immediate feedback. It's unlikely you're going to see any other industry get as badly affected because they're not the ones building AI. Lawyers are not optimizing KV cache and inventing new positional embedding algorithms at work because they can't do that shit like programmers can. Also, picrel btw. Only the US (where the massive financial circlejerk exists) is seeing any 'productivity', which in this case is measured only by output, not value of the output lmao.
>>110218691>Ed on so many occasions has said the tech is good and has improved. He has nothing wrong with LLMs. He expects local AI to win long-term and has no problem with that for that's not destroying the lives of millionsSo has he owned up to being completely wrong about LLM scaling and what their limit is? That's good to hear.>Feb 2024: "I believe we're reaching the upper limits about what generative AI can do and how accurate its outputs can be." >July 2024: "Generative AI, as I said back in March, is peaking, if it hasn't already peaked. It cannot do much more than it is currently doing, other than doing more of it faster with some new inputs" >Jan 2025: "I believe we’re at peak AI" >August 2025: "These models have clearly hit a wall where training is hitting diminishing returns" >Nov 2025: "the fact we're running out of high quality training data and we're hitting the walls of scaling laws, in the training paradigm, these models aren't getting better. What we're seeing today is pretty much what they're always gonna be like"
>>110218791He's absolutely right about the frontier labs not being profitable, and the AI industry hiding its debts via shady accounting and circular financing
>>110218791Kek I wonder what he thinks of Opus 5.5
>>110218791>2024>2025lmao I know how much it hurts you that he's had a solid 2026 streak, the year where it matters most. Keep holding on to your NASDAQ shares bro I"m sure it will pay off when we reach the singularity :')
>>110218818How good the models are influences their demand greatly. He has been consistently wrong in his predictions on how good LLMs can get. Thus his predictions about profitability are flawed.
>>110218888The demand doesn't outweigh the cost thoughbeit
>>110218888>How good the models are influences their demand greatly.When Gemini Carbon becomes available, all the chinks will distill it and have an Opus5.5-tier model. Will they be as good as the new Fable? No. Is Opus 5.5 intelligence but an order of magnitude cheaper enough for most people? Fuck yes. Gemini also have Xenon which could be Fable-tier and OpenAI have Bel. Anthropic and OpenAI can't thrive if the chinks have Google, Meta and Musk to distill (Musk has Cursor data now). It's a race to the bottom lmao. There is no moat and therefore no parabolic growth.
>>110218878>lmao I know how much it hurts you that he's had a solid 2026 streak, the year where it matters mostHow can you have a solid streak in predicting something that hasn't come to pass?
>>110218936btw, the Chinese AI-related tech index already popped, it's -40% from ath.
>>110218965
I've been reading Ed since 2024 but this year broke me. This stuff is crazy good.
>>110218429He is right and AI did peak in 2024.There's nothing as good as o1-preview and o3 was worse than o1-preview while both were removed. The 5 series is worse than 4o. Opus 5.5 and fable are only about as good as 4o was. The only advances have been with harnesses and provider-side system prompt (they prepend these to the prompt before the system prompt the user submits, openai renamed these system prompt and dev prompt recently but it's the same everywhere) because they can't improve the models themselves anymore.
There's no functional use case for AI that people would pay for other than really good porn, and people aren't going to pay more than $20/month for it.Zitron isn't wrong about AI being permanently unprofitable in a vacuum, he's wrong about it being permanently unprofitable in reality. The U.S. government is backstopping all of the companies and will be paying them tens to hundreds of billions of dollars each and every year via the Department of Defense. At most they'll let one or two companies (like Oracle) fail and prop up the rest forever. >but AI is so good at coding sloppaHow many workout trainer and calendar apps do we need? We had too many in 2022 and now we have 1,000 times as many. Nobody needs that crap and nobody is going to pay for it. People are only going to pay for porn and death dealing drones.
>>110215218isn't he the same bong that said browser cookies should be outlawed
> SIbot post
>>110219222I really hope he meant 3rd party cookies
>>110219210cybersecurity got destroyed due to AI, you can port large legacy codebases that would be multi year projects in a few weeks, gaming studios can port their old libraries to PC with a small team in a few weeks, QA can be replaced with Ai, it's great at low level optimization, and so on.... if none of that sounds "useful" compared to a lazy developer that drinks coffee for 5 hours a day, then you're either a chinese shill or unemployed.
For Anthropic, It's Pretty DisastrousThe Information reported last week that Anthropic has projected (made up) that it will make at least $12 billion in revenue in 2027, despite making $918 million in 2024 and losing $5.6 billion somehow.Anthropic has raised $14.7 billion to create an also-ran Large Language Model company that some developers like more than OpenAI, with a competing consumer-facing Large Language Model (Claude) that has an install base of maybe 2% of the five free-to-play games made by Clash of Clans developer Super Cell.Anthropic, much like OpenAI, has categorically failed to productize its Large Language Model, with the only product it appears to have pushed being Computer Use, a similarly-useless AI model that can sometimes successfully do in minutes what it takes you to do in seconds using a web browser.Anthropic, like OpenAI, has no moat. While it provides chain-of-thought reasoning in its models, that too has been commoditized by DeepSeek. Its models, again like OpenAI, are unprofitable, unsustainable and heavily-dependent on training data that's either running out or has already run out.t. ed zitronFeb 24, 2025https://www.wheresyoured.at/wheres-the-money/
>>110219210AI pr0n is total garbage right now as well. Literally unfappable if your IQ is over 65.The problem is AI really does have a billion super cool use cases, but it needs to pass the usability threshold which is basically a binary level. Under that level the usability is 0. Over that level it is a linear regime. Now we're still below the usability threshold, so it's only useful on demos: "but what if AI was yay good, wouldn't my demo be useful to you then?" is all it can achieve.>>110219275All heavily disproven bullshit.
LLMs will never be perfect, but you can engineer around that and make insane shit real.
>Fable 6 releases before IPO>Most intelligent model ever>Someone spends $10K building a non-trivial complex piece of software using it and it's great. >They start a company and sell the product. What's that company's value? $10K, because everyone has access to the same model, the same intelligence and can just create a clone in 24hrs, arguably it will cost less because the design, UI and functionality has already been thought-out by the other instance of Fable 6 that created it from scratch with the original creator, so it will be worth around $5K now just in raw compute/API spend.>original creator doomed>clonefag also doomed because their clone can be cloned>both end up losing money and have no value>both end up with no money to continue paying for Fable 6>Anthropic lose moneyThis is software in 2027+. Someone builds something? It's only worth as much as Claude or an uncucked model (which doesn't block reverse engineering) charges for compute to clone it. There are no software moats. Anthropic's largest customers are software companies.
>>110219321I fully believe that LLMs do not actually need to improve - in fact, we've actually overshot their required performance level - to achieve "insane shit". But that is because the craze to think LLMs must lead to AGI is insane. The correct solution is to use LLMs only as a small component inside a large engine. It will handle only the speech part, not reasoning, not math, not tool use, which will be relegated to similar large modules built for purpose. The real name of the game now is to optimize language models and make them smaller. Revive dNTMs or otherwise time-aware processes, revive model-driven state compression for infinite practical context size, and right-size the model for the capability we need. With no need for long-form reasoning and explicit hacks for specific topics and output formats, we can get very small, fast and efficient models. Separate code output and text output. Then create much bigger infra around the models to orchestrate them correctly, with fully interconnected layers (e.g. the code model can trigger a state change that causes the system to determine the next move is to call the text output model with instructions as to what to say next, which in turn calls back to the planning system when done to trigger tools needed for the next phase, etc.).None of that is magic, it's 2013-2016 techs with the extra powerful completion systems we have built. That would not only perform much better than current LLM slop but also run fine on commodity hardware.
>>110218591If you just ignore finance building a mansion for every bum on the street is also a great idea. The mansions are still useful after all. But, the problem is it can’t just be useful, it must be profitable, and, no matter what your favorite podcast grifter may have told you, profit is what’s left over after you pay expenses, including interest.
>>110218594Replacing every crud monkey on earth wouldn’t be able to pay back what they owe. Especially when you consider that by replacing them they lower their own value if they aren’t able to lock down big customers to their platform. Brute forcing math problems and writing broken portfolio websites is impressive technically, but it’s not a business model, that’s Zitron’s argument. Also, assuming you are a human being, why are you so desperate for everyone admit that LLMs are useful? Can openers are useful, should we invest all free capital in can opener factories and spend all our time listening to can opener influencers explain how the future of every industry is canning?
>>110219395> the craze to think LLMs must lead to AGI is insane.It's the only real reason financial capital and military funding are even there. Entire datacenters just for tools to be a small part of workflows is never gonna fly. It has to be grandly marketed the way it is because after AI big tech companies don't have any new maguffin to convince people we're making progress.It is beyond over when it becomes apparent AGI is never happening.
>>110219114
>>110218976SP493 soon to follow roody-poo
>>110219482>Technology that can solve hardest problems on earth like Navier Stokes >Not even the highest IQ individuals in society can solve them in 200 years>"Why do you keep talking about it? It's like can openers brooo!"The real question is: why do you copers keep denying reality?
>>110215242They forgot to include the last quarter in that graph, I wonder why?
>>110219526>hardest problems on earth like Navier StokesStill waiting on external verification for that one.
