-------------
Who is doing the research matters. What is presented here is not the product of academia. It's the product of a company that produces AI agents. The picture this web page paints may appear rosy and have just enough thorns to be convincing, but it's the equivalent of a tobacco company telling you that their product is neither addictive or carcinogenic.
I fully expect actual research will be done on the impact of AI and our hopes for it. This page, however, is marketing.
[1] https://news.ycombinator.com/item?id=47178371 [2] https://github.com/mickael-kerjean/filestash
However, since we're frank here, I'd say I'll download the most recent release and be very careful about upgrading because I don't put much trust in projects co-created with LLMs. I know there is a full spectrum but I've seen enough and I don't have the resources to check where on the spectrum your project ends up. LLMs are a powerful drug and terribly hard to stop once you start.
Also AI written, but I suppose that's expected. The big AI companies seem to want to make all their blog posts and communications have the AI tells so you know they didn't actually bother writing them
Investors want to see you use your own product, if they themselves don't feel the product is good enough to write their own announcement then investors would worry about their future.
And AI is still a product primarily aimed at investors and not consumers.
"I live hand to mouth, zero savings. If I use AI smarter, it may help me craft solutions to that cycle."
"Relaxing while my AI gets the work done, builds the wealth. It’s a shadow of me, just a very, very long one."
etc. I do believe AI currently accelerates businesses, especially in software dev. We work with a contractor who use Claude Code to reach incredible development pace for the size of their team, but also when we sit down with them in meetings they understand what's being created, they are able to argue their architectural choices, and they know how to propose business value.
You can't just buy a Claude subscription and have magically solve your problems. The thing is, as soon as Claude can do this without a business savvy human in the loop, then a) everyone can do it, so you won't actually have any value to propose, and b) Once the AI can run businesses without humans in the loop, you can bet your ass they will not out of the goodness of their hearts keep giving that ability away for $20.
In summary, AI if used to accelerate businesses _CAN_ be good. Buying it as a magic bullet to bring you out of poverty is probably a worse choice than just buying a lottery ticket.
That didn't last long!
Then there's people who are "well intentioned", I guess, but lack the technical knowledge. A friend of a friend with no technical background is selling websites to companies that he writes with Claude. They look shiny, everyone's happy in the short run, but I don't doubt issues will come up down the line that someone will have to be responsible for. I'd personally feel like I was ripping people off doing this, but I think also Dunning-Kruger prevents you from knowing any better if you are the type of person doing this.
Then there's the whole B2B SaaS gang that are basically just producing vaporware and telling other people how to produce more vaporware. This is no different from crypto, NFTs etc. before it really. Just people trying to hustle others.
And then there's the whole clawdbot gang probably burning more in tokens everyday than normal people use in a month so they can sort 18 e-mails.
So yeah I mean you're right, there certainly is a subset of people who are using this ethically (as ethically as you can use LLMs but that's another story) to make some money on the side. Certainly not the majority though I'd say.
The pseudo "entrepreneurs" who think they could outsmart the market by working less, are just naive. In a free market economy optimization is brutal and a freelancer developer will sell the same "product" cheaper, because he has the same technology available to him.
So the only way to get the gains from these AI technologies is to have something that can't be easily copied like market knowledge, data access or sweetheart deals with big companies that can pay more because their profits support the higher spend.
Also, services based SAAS especially B2B will not die, because a tyre shop won't have the time to write/debug/host it's own solution and will not want to depend to a single contractor who can disappear for a vacation. But the margins will go waaay down. 25$ for a set of forms and a database, not gonna cut it anymore.
True in the current state of LLMs, possibly not true forever if someone finds the magic bullet that turns the one-shotting (reliable) software dream that companies like Anthropic and Perplexity currently peddle into reality. Seems far-fetched ATM but the gains since GPT-2 have been very real.
We're quite a ways away from this though, even with Opus 4.6 and the like. And even further from it being part of Claude Code rather than some proprietary $1000/mo. closed-source solution.
As you say though, _if_ such a technology were to exist, it's Anthropic that holds all the cards, not random entrepreneur #25721 who is asking the Anthropic API the same thing that the actual customer could just be asking directly. At that point you're an undesirable middleman, not a business.
The quote about being temporarily embarrassed millionaires comes to mind….
Louder for those at the back.
> You can't just buy a Claude subscription and have magically solve your problems.
If Average Joe were to delegate most of their life decisions to the chatbot, it'd probably turn out better, or in the worst case, more informed.
Okay how are we measuring this? We can't even quantify intelligence for humans accurately, let alone compare it to machines. Hell, we can't even really define intelligence.
I mean, humans can learn on the fly and progressively, and currently no LLMs are capable of that. Literally none of them, and no context doesn't count. So if that's the measure, then LLMs sit at a 0 along with rocks and twigs and humans closer to a 1.
Obviously that's not really the measurement, LLMs are quite good. But I don't think we can say, for sure, LLMs are a replacement for humans. They might replace some specific tasks, but humans are not a set of tasks. I'd still rather have 10 engineers than 0 engineers and 10 Claude Code licenses.
Even if true, no one is ever going to do that because of Dunning Kruger, so it's still not magically solving problems.
Fake it till you make it mentality that degenerated completely once we got the internet. It used to be "crypto will make you rich, buy my coin/course", now it's "AI will make you rich buy my tool/course", the same type of people will get fleeced
These are the people getting all the attention: https://www.youtube.com/watch?v=NwaUMBQ3Wgg
Some quotes that stuck out to me:
"I’ve been working on a scientific project for 6 years... with Claude I was able to accomplish in 5 weeks what took me 6 years. I’m old... I estimate I have another 5 to 10 years and I’ll accomplish everything I want." Academic, Germany
"I live in a war zone... AI can not only give practical advice, but also emotionally calm me down during panic attacks. It can calm someone during a missile attack in one chat, and laugh with me about something silly in another. That’s what makes it not fragmented into a therapist/teacher/friend, but something whole." Ukraine
"If an AI had been in Stanislav Petrov’s position — the Soviet officer who prevented a potential nuclear war in 1983 — it would not have refused to launch." Academic, USA
"The humans in my life were telling me it was psychological. An AI chatbot was the only one who really listened and took me seriously — it pushed me to ask for specific tests... which came back 6 times higher than its supposed to be."
I can see this kind of survival-bias stories distorting the reality. To have millions of people asking for "specific tests" because AI told them seems problematic. One in a million will discover something, and that story will be enough to create the believe that is "worth doing the test that AI says" just in case. But...
> which came back 6 times higher than its supposed to be.
It has been proven that massive testing creates many false positives.
This happened during covid: https://www.bmj.com/content/373/bmj.n1411/rr
Tests may not be as reliable as though but they are good enough when other symptoms are accounted for. To randomly test people based on AI hallucinations can increase the number of unnecessary medication or even interventions.
This is a competition of public and private interests. A sick individual is going to lobby for tests until they discover the cause. From a public perspective, it might be cheaper to just let them die. AI is an advocate for the individual.
For the record, ChatGPT helped me diagnose a lifelong illness. I'm a new man now thanks to AI. Literally life changing. I had spent decades pleading for tests because no one could figure out the cause. I think a likely outcome here is not necessarily 10,000x more tests performed, but similar or even fewer tests, because the diagnosis success rate with AI is higher. It's not subject to bias. People tend to be more honest and reflective with their AI than they are with doctors. They get 5 minutes to give the entire case to the doctor. With an AI they can spend weeks debating and reflecting. This builds a case history far more detailed and accurate than anything we have in modern medicine today. Amplified by an order of magnitude because the AI can extract meaningful insights from the discussion.
In the very near future our AI will contact our GP for us. Soon after that, our GP will be our AI.
They're good at acting as a "reverse dictionary" like this where you give it a description of something, and it knows the word for it. They have approximate knowlege of many things.
That was my take with the entire report which I think lends to an inherent bias within the data and stories. You have the entrepreneurial stories, then you have the ones where people are both impacted and receiving benefits.
The infographics and charts even call out how countries that are "first-world" with fewer safety nets are more likely to be in "survival" mode compared to countries with them.
The bit from George Carlin standup routine regarding how the poor are there just to scare the hell out of the middle class rings true in this reflection. Poorer countries accept their current realities and the feedback reflects the hustle. Richer countries with safety nets reflect the existential issues with previous industrial revolutions. Richer countries without safety nets reflect the fear that their efforts will be made "replaceable" by AI.
As for the rest - massive testing creating false positives - that is an issue of implementation and the errors introduced by humans, not data itself. If the process were in large part made more automated, it could screen for a larger panel of issues for less cost.
From my experience working deep in data and human factors - the issue in quantifying the root cause isn't reality, we live a shared experience in general. The issue is the data isn't good enough. What bugs us about it is the psychology that our perceptions are different enough to the degree that we will fight to prove an unknown.
The fact I'm young-ish and healthy looking, with good skin and hair, leads many doctors to outright dismiss me. Never mind my history of cancer and the undeniable fact that I am obviously not healthy. But I can also use the squishyness to my advantage. I talk confidently, I push back, and that works. It sort of short-circuits a lot of doctor's brains.
> "It’s not healthy to love someone or something that can’t tell you no." - Not Currently Working, United States of America
> "Instead of AI doing my chores, AI does the stuff that I love—in two minutes, without any passion." - Student, United Kingdom
> "I used to write songs for my kids. Now I have [AI music product] make them for me. I used to write poems for those I loved... I used to bust my brain doing research, and now I get a research summary that is better... but I didn’t learn the paths in between. And yet, I use it because I have to pay off my house, pay off my land, and feed my little kids so I can find an hour on Saturdays to do something meaningful with them." - Software Engineer, United States of America
> "I believe AI is likely to kill me and everyone I love… building an AI that’s smarter than us before we’ve figured out how to keep it under control will likely destroy everyone and everything they value." - Software Engineer, United Arab Emirates
This was one of the highlighted quotes:
> "I’ve been told I’m ‘too much, treatment resistant, complex’ by providers. Within six months of working alongside AI, I was able to understand my own inner world in a way I never could before. I was doing creative writing again after quitting for two years. I developed hope again — that’s the through line." - Healthcare Worker, United States of America
A healthcare worker outsourcing their own treatment to an LLM, who won't tell them no, is terrifying.
There's always something about claims like this. I'm not claiming that AI can't speed up your processes, but I question the persons expertise when they claim months or years of work turns into days or weeks. It just doesn't make sense to me.
"My output is like 25x what it used to be. I’ve built over 20 backend server tools, 7 major projects in the last 6 months—my work output this year is greater than the last five combined. I can typically finish a significant project in a day or two."
I am not sure if this would be true given how AI's have refused to kill processes.
If AI is programmed to always serve its makers as some are arguing, then it would certainly become true.
For the record, Petrov made this decision based on a false assumption that the US wouldn't launch just a few missiles, but would instead send a lot, all at once. Except, that one of the US plans was to send a few missiles to destroy critical targets, and then follow it up with a large scale attack.
Petrov himself said that he might've acted differently if he was aware of this possibility. And even then, his initial hestitancy was basically a 50/50 gamble.
An AI would basically do the same thing if asked - just roll a random number, and launch nukes below a threshold, adjust threshold based on some llm evaluation of the situation if needed.
The coming years will see the current RAM shortage followed by a war between local AI models and vibe-coded shitware “productivity” software for memory on our devices. Especially fun will be when vibe coding crap hits corporate security software, which is already often so bad it looks more like sabotage than security. Imagine when it gets, from both angles (using models for threat detection; vibe-coded shitware) another large multiplier on its resource use.
Then your comment made me curious, and I clicked. What the actual fuck.
Marketing and entertainment are supplanting news and knowledge. I hope that the people that is pushing back succeed.
Stationary gasoline engines were already changing the farm and reducing the head of horses necessary to feed a nation. It, too, was a faster horse for them.
Anyways.. it took the Detroit police to eventually deploy the first automatic stoplight. The real innovations seem to be often found downstream of the simple increases in capacity.
That all being said, it seems to me the current crop of LLMs haven't done this, their power and training budgets do not seem to be scaling favorably against adoption rates and profit margins. Absent a significant change in algorithm or computing substrate I don't think this strategy is the leap everyone hopes it will be.
The tension is real though. The #1 concern (unreliability at 26.7%) maps exactly to the problem with how most SMBs use AI today — chatbots for general Q&A rather than purpose-built tools for specific operations. A chatbot that hallucinates an answer is annoying. An invoice processor that hallucinates a number is dangerous. The architecture has to be different.
Wrote up a deeper analysis of what the findings mean for small businesses: https://falaah.ai/blog/what-80000-people-want-from-ai/
> AI should learn to say two things: ‘I don’t know’ and ‘you’re wrong.’
My guess is, the next evolutionary step of LLM's should be yet another layer on top of reasoning, which should be some form of self-awareness and theory of mind. The reasoning layer already has some glimpses of these things ("The user wants ...") but apparently not enough to suppress generation and say "I don't know".
https://petergpt.github.io/bullshit-benchmark/viewer/index.v...
"The doctors were just doing a copy-paste of a copy-paste of a prescription from a few weeks ago, not realizing it was the medication that was killing her. AI helped me ask the right question to save her life."
What consumer benefits is ai driving? at least with industrial automation consumers benefited from new technologies, cheaper goods, and new job categories.
It would be great if there was some internal “make this benefit Main Street and knowledge workers” department, helping find ways for workers or creators to capture the value of some of the increased productivity.
If they wanted to do this, they could put their models in a public trust for the public's access and benefit in research, education, etc. Then it could be licensed, pay a dividend like a sovereign wealth fund, etc.
Considering that they copy and train on the sum total of all human creativity, a public trust is something that would be in line with both the spirit, and first and fourth considerations, of fair use doctrine:
1. the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;
2. the nature of the copyrighted work;
3. the amount and substantiality of the portion used in relation to the copyrighted work as a whole; and
4. the effect of the use upon the potential market for or value of the copyrighted work.
What I need instead is something that takes the burden off my entire society and gives them a breather. Universal health care to start. They could also use a higher minimum wage, and lower housing costs.
You'll either need to freelance, or start a company (or maybe a co-op) to capture the new value created by your ability to leverage AI.
It won't be much different to when a company buys more CNC machines and the employees don't get any more money despite producing way more parts.
This is quite easy. Just optimize the models to do reviews and bug finding. This would make developers (who normally hate reviews) quite happy and let them do more coding, thus delivering more value and possibly earning more...
I have no clue what this would look like other than maybe an investment fund for people creating apps/businesses based on Claude tools.
The intrinsic satisfaction of increasing the wealth of shareholders. We should all be happy to devote ourselves to getting them more, nothing is more important than that.
My kids like to use AI to discuss things they learned in school in greater depth, and from different angles than they learned in the textbook. They can also ask "What if" and "Why not" questions from this infinitely patient teacher.
AI chat bots will summarize the top N web search results as if they're fact, weaving them into seemingly coherent narratives, all while reassuring the user that their questions are really good and they're learning a lot.
Basically consumers don't really pay for software in the first place, and the leverage from labour companies get through software is already through the roof even before AI. Will much change for consumers of software?
So... not much benefit either.
But this implies higher productivity, no? This must mean more outputs that should benefit someone, unless the jobs that are being automated had little value to begin with. Seems paradoxical.
The number 1 ask from the interviewed cohort is « professional excellence »
It is telling about what we prioritize in our society.
I am usually an optimistic person, but I struggle to see how this does not end up with more misery and worse lifestyle all around.
No it's not.
It's a measure of what people want accomplished, are least interested in doing themselves, and feel capable of reviewing enough to delegate to a known lair.
To remove some of that bias, I'd recommend to get an independent body (probably some university) in and let them do the interpretation and write the article.
I just want people to see the tactic for what it is. I really like Claude Opus 4.6 but this just screams "marketing" to me. I wouldn't say it's wrong, it's good to have these discussions and I'd encourage AI companies to say what they have to say. I would say: more independent sources are needed (and not another AI company).
I can tell you the questions are biased from the start. That study has to be redone entirely.
The quotes they have are really interesting to read. That's what I was hoping to get when I built mine.
And just keep scrolling, you can make it to the story eventually.
I'm not going to claim I know this response was written by an AI, but it's very suspicious. I would like to hear about how Anthropic ensured that the survey responses were provided by real human beings using their own words.
Why not call it a "double-edged sword" or something else? Light and shade are opposites but not necessarily two products from the same tool. It just irks me.
4.0 MB transferred
I'm much more curious about the results of 80k people who don't use AI regularly.
01. Professional excellence 18.8%
02. Personal transformation 13.7%
03. Life management 13.5%
04. Time freedom 11.1%
05. Financial independence 9.7%
06. Societal transformation 9.4%
07. Entrepreneurship 8.7%
08. Learning & growth 8.4%
09. Creative expression 5.6%
I find this highly suspicious. I'm sure there would be at least 10% who respond "I want it to go away".
> These are active Claude users who'd already found enough value to keep using AI, and our interview asked first for positive visions for AI and then for concerns that would counter their vision.
It's like those recipe sites that have 5 pages of nice photos and background story and side tracks and whatnot as the author waxes verbose, so they need to put a 'Jump to recipe' button in so people don't just click 'Back' immediately.
Except this time for an article.
I can't tell if 'skip the junk' is good (junk can be skipped!) or bad (maybe this means there's too much junk on the page?)
The meme that's been going around that em dashes are an indicator of LLM output is completely invalid, and people who repeat it are really just outing themselves as the sort of people who don't actually read very much.
Is this incompetence or a deliberate error to indicate human authorship?
If the former then why aren't they using at least an AI to proof read? If the latter then what do anthropic think is wrong with AI written text?
Who writes this unrealistic gibberish?
Hell, anyone who’s been interviewed by a journalist can tell you they do it, too, sometimes to the point of changing the substance of what you said.
Not 81,000 as it says in the title. I know I'm being nitpicky, but I wouldn't round up to 81k. Surely the 'important number' in this case is 80, so you would round down to that. Then let the reader pleasantly discover you had interviewed ~500 more than you stated.
It's funny to me when someone does this sort of minor hyperbole that's verging on lying - you have to wonder what is going on.
What would you have done, in those circumstances though? Would you 'round up', overstating your case, like Anthropic did? Or would you ensure that you avoid the suggestion that you were misrepresenting the numbers?
> I'd started telling Claude about things I couldn't even tell my partner. It felt like I was having an emotional affair.
> I worked with an AI to prepare educational materials for my eldest child—asking the AI to work as both tutor and curriculum expert. We received [my child’s] report yesterday, he was graded as either ‘Above’ or ‘Well Above’ standard in every academic area he studies.
So many concerning quotes that read like AI is a workaround for things that are not working right at scale in society.
I worry (1) AI workarounds will make it clear society can tolerate even more suck then (2) society will get worse to where AI is required to cope then (3) AI will stop being subsidized and the poor will get wrecked.
"Electric lighting means I have time to read books." - omg society made people work during daylight! (This is a real issue in Africa btw, or at least was not very long ago.)
If AI leads to better medical diagnoses, great!