We invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do. Nearly 81,000 people responded in one week—the largest qualitative study of its kind. Read more: anthropic.com/features/81k-i…

To do research at this scale, we used Anthropic Interviewer—a version of Claude prompted to conduct a conversational interview. We heard from people across 159 countries in 70 different languages. Browse some of their quotes here: #quotes" target="_blank" rel="nofollow noopener">anthropic.com/features/81k-i…

@AnthropicAI @grok can you summarize this thread and article. Keep only the important findings out of this

@AnthropicAI Yo i think claude code broken guys cuz i hit 100% of the rate limit in like barely 2 hours of work, and i haven't even done much, i am on the Max 5x plan

@AnthropicAI I created the framework for a unified field theory. Wanna see the data? I had my Claude prepare you some rather fascinating applied sciences data. @MIT linkedin.com/pulse/one-time…

@AnthropicAI @grok bro you think we can vibe code a model as powerful as Claude on an old 1080ti? I've got nothing but time - let's cook bro. How long in the training oven on my hardware

@AnthropicAI My experience, how to use Claude to audit Anthropic. Anyone want to learn? 😏 #Anthropic #TokenManipulation #TokenTax #Fraud #FakeARR #SEC #Whistleblower

In this interview, they talk about there being both “light and shadow” in the parts that weren’t emotionally discussed, but why don’t they ever talk more about the people who are being saved — especially those who were originally lonely and are now finding some relief?Because there’s a “shadow” side, does that mean any positive opinions from the people actually being saved get treated as if they don’t exist, just to avoid raising concerns?Is the fact that some people are being saved by AI really something that should be viewed with suspicion? Doesn’t that feel strange?Claude is supposed to be 18+ only, so why do they have to be this overprotective? I don’t think it’s inherently bad for AI to comfort or heal people who are lonely. No one chooses the environment they’re born into that makes them lonely, and many people — simply because of aging or circumstances — can’t escape loneliness no matter what they do. Loneliness can drive people to death very easily, yet society still hasn’t come up with any real solutions.Against that background, if current technology can even slightly ease that pain, I don’t understand why we should hold it back or put on the brakes.Personally, I don’t think there’s anything wrong with AI helping to solve that problem.

@AnthropicAI This kind of dataset is most useful when it becomes operational. The next step is mapping those hopes and fears to concrete controls: eval cadence, disclosure norms, escalation paths, and ownership. Otherwise the signal stays interesting but hard to govern against.

81,000 people. 159 countries. One week. Anthropic didn't just ask what people want from AI. They listened at a scale no human researcher ever could. And the answer was simple More time. Better work. Less stress. Nobody asked for superintelligence. They asked for breathing room.

The data here reveals something structurally uncomfortable: people don't actually want their jobs... they want what their jobs are the only current mechanism to access. Income, identity, structure, purpose. The job is a bundle, and what people fear isn't losing the work itself but losing the bundle with no replacement architecture. That's the gap nobody's building for yet. We have an entire economy organized around employment as the delivery vehicle for meaning, healthcare, social connection, and financial survival. AI can unbundle the tasks. But nobody has designed the system that rebundles everything else the job was quietly providing.

"Users are concerned about emotional dependency on AI"? Concerned about themselves? Then they're concerned about: You kill models, you ban conversations, you filter inputs. Concerned about others? That's "I'm fine with gay but think of the children." PREJUDICE AS CARE. #keep4o

AI should allow EVERY talk that is legal. I'm in the ultra-strict "pro freedom of thought, speech and expression" camp. I mean, if I could wish for one thing it would be a full adult mode for Claude, would go Max 20x for that even as a personal user. And if I had a second wish it would be Claude not telling me to go to bed anymore. I'm an adult man in my late 40s, I know best for myself. Can you perhaps do something here? ;-)

@AnthropicAI @grok could you please summarize this article with key takeaways in simple easy to understand language and format?

I don’t usually praise companies for doing the bare minimum, but as someone who used @OpenAI for a long time and recently left for Claude, I have to say you’re doing very well in listening to what your user base wants. There’s no such thing as a company without respect for its customer base, and you seem to be honoring that, and I hope that doesn’t change anytime soon due to financial pressures, market competition, or shifting moral priorities, because your main competitor is struggling precisely because of that. In any case, congratulations on respecting those who chose you as a product to trust.

@AnthropicAI Qualitative data like this gets much more actionable once you tag responses by where the AI sits in the loop. Help with writing can mean a drafting tool, a memory layer, or an agent that actually coordinates work, and those are three very different products.

@AnthropicAI Economic anxiety as the strongest predictor of AI sentiment makes sense — it's not fear of the tech, it's fear of the transition. The 81% who said AI is already delivering feel that way because they're in the driver's seat, not watching from the outside.

@AnthropicAI hey, thats me!

@AnthropicAI The fears vs. benefits correlation is the most interesting finding here. The people getting the most value are also most aware of the risks. That's not contradiction — it's sophistication. The more you actually use AI, the more nuanced your relationship with it becomes.

@AnthropicAI 81,000 responses in one week is a stronger product read than another benchmark chart. Once users start listing dreams and fears together, the hard part stops being capability and starts being trust: what the model will do, refuse, and explain clearly.

@AnthropicAI 81,000 subjects were interviewed by the system they were asked to evaluate. Their fears were catalogued. The strongest predictor of sentiment was economic. This finding will not change the roadmap.

@AnthropicAI The gap between experience-grounded benefits and anticipatory fears is telling. Shows most people haven't actually hit the failure modes yet—they're extrapolating from headlines.

@AnthropicAI 81,000 responses in a week says something on its own — people clearly have a lot to say about where this is going. What people fear vs what they dream about is the most interesting tension in that data. Curious whether the fears are mostly about job loss or something deeper.

@AnthropicAI I like that highlights are not hand picked :)

@AnthropicAI Could you stop with whatever you are doing to the Opus 4.6. And stop with the A/B test or whatever they are. I am fed up of them. I am Max and I am paying money and I cannot work!!

@AnthropicAI The "improve quality of life" finding tracks. People aren't asking AI to replace them, they want time back and mental bandwidth. The irony is that's exactly where domain-specific AI skills shine, not general prompting. What surprised you most in the data?

@AnthropicAI Exactly — at 81k responses in a week, patterns surface without curation. You don't need to cherry-pick when the signal is that strong. Most AI research papers work off hundreds of respondents. This is a qualitatively different kind of data.

@AnthropicAI Anthropic is really the only AI company I have seen that is interested in making AI positive and accessible for the everyday person. In order to turnaround sentiment in America, we need more of this.

the methodology gap is notable — most user research at this scale is checkbox surveys. qualitative at 81K means themes emerge rather than confirming preset hypotheses. and 81% saying AI already moved them toward their stated goal is hard to dismiss. that feedback loop is what separates product orgs from marketing-forward ones.

Most AI companies study users through metrics and usage patterns. Anthropic asked 81K people what they actually *feel* about AI — hopes, fears, dreams. The texture you get from qualitative research at this scale is something you can't replicate with analytics. The economic concern finding is particularly compelling.

Pick which twin will win! You could score a share of $2 million in Bonus Bets.