AI with Kyle Daily Update 163

Anthropic interviewed 81,000 people about AI.

Discussed at [00:00:34]

Anthropic used their own AI interviewer tool to conduct qualitative interviews with 80,508 Claude users across 159 countries in 70 languages. They're calling it the largest multilingual qualitative study ever conducted.

Here’s the full study - very worthwhile: https://www.anthropic.com/features/81k-interviews

Quick caveat - these are all Claude account holders. Not the general public. So the data skews towards people already comfortable with AI. Still useful though!

The top thing people want from AI? Professional excellence (18.8%). Making work better, automating the boring bits, freeing up time for the stuff that actually matters. That's exactly the use case I push for.

Second highest? Personal transformation at 13.7% - using AI as a coach, guide, or for mental health support. Companionship is in there too, which lines up with what we saw from the Open Router study where role playing and companionship was about 40% of all token usage.

Entrepreneurship came in at 8.7%. Doesn't sound massive, but that's nearly 10% of Claude users wanting to build businesses with AI. Way higher than the general population's entrepreneurial drive.

On the worry side - unreliability (hallucinations, fake citations) is still the top concern at 26.7%. I'm surprised it's that high honestly. With search modes, deep research, and chain of thought, this has got massively better in the last year. If you're still hitting hallucinations regularly, you're probably on a free tier or an older model.

Jobs and economy is second at around 20%. Here's the uncomfortable truth people don't want to hear - every previous technology created new jobs to replace the old ones. The difference this time? AI can also do the new jobs it creates. We're not the only option for labour anymore. That's new.

One quote from the study stuck with me - "AI should be cleaning windows and emptying the dishwasher so I can paint and write poetry. Right now it's exactly the other way round." Pretty much sums up why people feel weird about all this. It’s hitting the white collar workers first. The people who thought AI would help them the most.

Interesting regional split too. Sub-Saharan Africa, Latin America, and Southeast Asia are far more positive about AI. Western Europe and North America are more concerned. Makes sense - if you're in an economy that's been historically “on top”, a technology that levels the playing field is scary. If you're in a developing economy, it's the biggest opportunity you've ever had. Exciting!

Kyle's take: Only 11% expressed zero concerns about AI. Those people and the people who think AI is pure evil - both groups haven't thought it through! There are real problems with AI that can’t be ignored.

Most of us are somewhere in the middle, grappling with something that's genuinely good AND genuinely dangerous. If you think it's all black or all white, you're probably not thinking hard enough.

Claude Dispatch: Open Claw for grownups is taking shape

Discussed at [00:23:24]

More big news from Claude this week. Dispatch is a new feature in Claude Cowork that lets you talk to your always-on computer from your phone.

So what? well, if you've set up Cowork on a Mac Mini or a laptop that stays on at home, you've got this persistent brain that knows your business, has access to your files, and can run scheduled tasks. Now you can message it from anywhere. Walk to the shops, send it a task from your phone, come back and it's done.

I've been saying for weeks that Anthropic are basically rebuilding OpenClaw inside their own ecosystem. Persistent memory - added. Remote control - added. Cowork - launched. Now Dispatch. They're ticking off the features that made OpenClaw exciting, one by one. And I’m here for it.

The trade off is you lose some of the wild flexibility OpenClaw had. You can't go completely buck wild with connections. But you get way more security. Remember the OpenClaw horror stories - people getting their inboxes deleted, hard drives wiped? That should happen a lot less in the Claude ecosystem. Notice the less here. It’ll still happen!

Kyle's take: When Openclaw came out, I told most of my audience to sit it out if they weren't technically comfortable. That was one of the very few technologies I've said that about. Normally I’m super encouraging of everyone at least trying! But Openclaw was fine to miss because I knew the big labs would be making something similar.

And here it is!

With Claude Cowork and dispatch, we're getting close to the point where anyone can have this kind of always-on AI assistant. That's a big deal. I expect Claude Chat, Cowork, and Code will eventually merge into one product. They're clearly heading that way.

Why Anthropic doesn't care about your $20 a month

Discussed at [00:33:17]

A lot of ChatGPT refugees are moving to Claude right now. And they're hitting the same wall Claude users have dealt with for years - usage limits. Use Opus on the $20 Pro plan and you'll run out after a few exchanges.

I keep seeing posts like "Dear Anthropic, the ChatGPT refugees are here and here's why they'll leave again." People are writing heartfelt open letters begging for better consumer pricing.

Here's the thing - Anthropic don't want those users.

@iamkylebalmer

Anthropic's API revenue beats OpenAI. Their consumer subscriptions are tiny. If you're on the $20 plan you're costing them money. Here's w... See more

Look at the revenue numbers. Anthropic's consumer subscriptions are tiny - under $1 billion compared to OpenAI's five and a half billion. But their API revenue - selling to builders and businesses - actually beats OpenAI. That's where the money is. That's what they're building for.

If you're on the free plan or the $20 plan, you're costing Anthropic more than they make from you. They want people on the $100 or $200 plans (which are still subsidised!), or ideally using the API, where the actual money is,

And here's what's coming - OpenAI is noticing this. Their head of product told staff they can't be "distracted by side quests" while Anthropic gains ground in enterprise and coding. OpenAI are shifting towards business customers too. Which means ChatGPT will probably get worse for free users. More ads. More restrictions. The $20 plan getting squeezed from both sides.

Kyle's take: I'm on the Max plan and I still hit limits occasionally. That's the reality of where this technology is right now. If AI is central to how you work or build, budget for the higher tiers. If you're using it casually, the free tiers will get rougher but they'll still exist. Just don't expect these companies to optimise for you as an individual.

Kyle