You’re probably wondering if you rely on AI a little too much. Good news. The thing you rely on most has something to say about it.
Anthropic’s chatbot, Claude, now has a “Reflect” feature. It’s a mirror held up to your own usage habits. Not in a judgmental way. Well. Maybe a little. But the point is introspection.
Anthropic claims this tool helps you answer big questions. Are you using the bot effectively? Are you overusing it? Are you outsourcing tasks that you really should handle yourself?
The company isn’t shy about calling out cognitive offloading. We worry about atrophy. The fear is real. We hand over too much thinking to the machine and forget how to do it ourselves. This dashboard is their counter-argument. A nudge to check your balance.
The numbers game
The insight dashboard shows you the cold, hard truth. Or at least the data.
- Key topics you keep circling back to
- Usage patterns over time
- Frequent tasks where you dump the work
You can look at the last month. Six months. A year. It tracks it all. There’s also a feature for quiet hours. You set them. You can dismiss the nudges to take a break, sure. But the prompt stays.
“What’s one thing you want to do yourself, even if the bot is faster?”
It’s a direct question. You have to answer it. You talk it out with the AI. It forces a pause. Step back. Examine the role this thing plays in your life. That’s the goal.
What gets hidden?
Privacy still matters. At least some of it.
The insights don’t touch incognito chats. They don’t dig through connected files in your email. They ignore health data linked to other apps. You can breathe there.
But here is the catch. Reflection insights do cover sensitive conversations. Mental health discussions? Yes. But only at a high level. Vague summaries. If the bot shared a helpline resource before, that might show up too. Just so you remember you looked for help.
Anthropic says they worked with independent experts. They wanted to build tools that actually help people figure out what works for them. Not just more data. Better questions.
It’s odd they’re asking you to think critically about the tool while the tool does the prompting. Ryn Linthicum, head of wellbeing policy, notes they consulted youth development experts too. Even though users must be 18+, they want to help parents understand how their kids are using AI.
It’s currently in beta. Free for paid subs with memories on.
The tool works. The data is clear. But when you sit there staring at the stats, wondering if you should type more or think more… who decides what the right answer is?
Maybe that’s the real point.
