What 81,000 People Actually Feel About AI (It's Not What the Polls Say)
Most public opinion polls say AI sentiment is cratering. Anthropic just ran the largest qualitative study of AI attitudes ever conducted, 81,000 people across 159 countries in 70 languages, and found something more honest: most people are not choosing between hope and fear. They’re carrying both at the same time.
That feels true. It feels true because it is the actual human experience of this moment.
The Main Story: What 81,000 People Actually Told Claude
Anthropic built a special version of Claude designed to run open-ended interviews. In one week, it conducted 81,000 conversations across 159 countries in 70 languages. The goal was simple: ask people what they hope for from AI and what scares them.
What happened: The study found that professional excellence was the top-reported hope. People want AI to help them do better work, free up time, achieve financial independence, and manage life more effectively. The top fear was AI getting things wrong, followed by job anxiety, loss of personal agency, and over-reliance. Sentiment varied by region: India and South America ran above average, while the U.S., Europe, Japan, and South Korea ran neutral or below.
Why it matters: The mainstream poll narrative, that AI favorability is collapsing, misses the nuance. People are not anti-AI. They are holding complexity. They want what AI offers and they are worried about what it costs. That is not opposition. It is the reasonable posture of someone paying attention.
The TWO angle: Almost as interesting as the findings is the method. Claude ran 80,000 in-depth interviews across 70 languages in a single week. The technology being studied was also the instrument doing the studying. That is a remarkable capability. It is also just a capability. “For what will it profit a man if he gains the whole world and forfeits his soul?” (Matthew 16:26). The people in this study are not data points. They are image-bearers with real hopes and real fears. The tools that help us understand them faster do not replace the obligation to treat them with dignity. AI can scale listening. It cannot scale love.
The Rest of Today
Cursor shipped its own coding model, and it competes with frontier at a fraction of the cost. Composer 2 outperformed Claude Opus 4.6 on the independent Terminal-Bench 2.0 benchmark and sits within five points of GPT-5 on Cursor’s internal tests. The price: roughly 1/10th of GPT-5 and 1/20th of Opus at comparable speeds. Cursor went from using other companies’ models to building one of its own that competes with the best ones. That is a meaningful shift in the economics of AI pair programming.
Google rebuilt AI Studio into a full-stack app builder. The old AI Studio was a prompt playground. The new one pairs a coding agent called Antigravity with Firebase’s backend, where user accounts, data storage, and authentication are set up automatically. The roadmap includes Google Workspace integration, one-click deployment, and connections to payment processors. Google and OpenAI reached the same conclusion on the same day: the era of separate AI apps is ending, and both are racing toward a single surface that handles everything. For vibe coding specifically, this is the most accessible full-stack option yet for non-technical builders.
Perplexity launched Health for Pro and Max subscribers. Users can now sync wearables, lab results, and medical records to create personalized dashboards, ask questions about their own health data, and cross-reference it with medical journals. Separately, Perplexity’s Comet browser is now available on iOS as a free download. The health data integration is the more interesting move. It turns Perplexity from a general agent into something genuinely personal.
Microsoft’s MAI-Image-2 landed at No. 5 on the Arena.ai image leaderboard. The biggest jump from its predecessor: text rendering, up 115 points. That makes it genuinely useful for posters, slides, and infographics. It is free to try in Microsoft’s MAI Playground for U.S. users. Microsoft has been signaling its intent to reduce reliance on OpenAI and compete with its own models. MAI-Image-2 is the clearest evidence yet that the signal is real.
Jeff Bezos is reportedly raising $100B for Project Prometheus to buy and automate aerospace, chipmaking, and defense companies with AI. The fundraising included stops in the Middle East and Singapore. This is notable less for the dollar figure, which is large enough to lose meaning, and more for what it signals: serious capital is moving toward applying AI to physical world infrastructure, not just software.
OpenAI acquired Astral, the open-source startup behind Python’s most popular developer tools (Ruff and uv), folding the team into Codex. The same day OpenAI revealed it is merging ChatGPT, Codex, and Atlas into one desktop superapp. The major labs are consolidating their surfaces. That pattern is now consistent enough to plan around.
One Tool Worth Knowing
Claude Code Channels (research preview): Anthropic released the ability to message your Claude Code session directly from your phone via Telegram or Discord, with Claude messaging you back. If you have ever started an agentic coding session and then had to walk away from your computer, this matters. It means the session does not have to wait for you. You can check in, redirect, and get updates from wherever you are. Early and rough, but the direction is clear: coding agents that work while you are away and stay in touch while they do it.
There is a high school student in the Philippines, a class secretary, who spent a week building a real website with ChatGPT when his class president expected a Google Drive folder. He had never written HTML. He built it anyway. He is now exploring full-stack development.
That story sits underneath all the benchmarks and funding rounds and superapp announcements. The barrier between “I have no idea how to do this” and “I built the thing” keeps getting lower. But lowering the barrier is not the point. God gave people the ability to create because He is a creator, and we bear His image (Genesis 1:27). The question is not just what you are going to build now that you can. It is whether what you build honors the One who gave you the impulse to build in the first place.
Comments
Back to all posts