The data backs up what I keep saying
Hey friend. In almost every workshop I run, someone over 50 tells me they're worried they're too far behind on the tech to get much out of AI. This week Anthropic put out a study that says close to the opposite, and I couldn't wait to share it with you.
Here's what I wrote about it this week, because the numbers reflect the people I work with every day.
Anthropic just proved it with data: your domain expertise matters more to your AI results than your technical skill with the tool.
That's the core finding from Anthropic's study of 400K Claude Code sessions. Experts consistently outperformed novices. The reason wasn’t because they knew the software better but bc they knew their subject better.
Experts hit verified success rates of 28-33%. Novices hit 15%. Verified success means the work actually got done, not just that the AI responded.
How?
Experts bring more context to every session. They ask better questions, give richer instructions and get deeper outputs (12 actions and 3.2K words per prompt on average, versus 5 actions and 600 words for novices). And when things get hard, they stay with it. Novice sessions were abandoned nearly one in five times (19%). Experienced users almost never walked away (5-7%).
That gap comes from knowing what good looks like. You've seen enough in your career to recognize when something falls short + you know how to push for better.
And there was another surprise -- on pure coding tasks, lawyers, managers and scientists nearly matched software engineers!
How did they perform well at coding -- judgment and context and decades of knowing how to think through hard problems, imo.
That's exactly what you have.
If you're over 50 with deep domain expertise, using AI effectively is absolutely possible. You can use AI to multiply your experience, which is pretty powerful I think.
So if you've been waiting to feel technical enough before you really dig in, the study says you already hold the part that's hard to teach.
Free Lesson: Personalize Your AI So It Finally Gets You
Join me in a free, hands on session to customize your AI to give you better answers. You’ll build your own context file in 10 minutes, together with the rest of the class and supported by me. You’ll walk away with your ChatGPT, Gemini or Claude set up to work with your background and goals every time.
Sign up even if you can’t make the day & time, and you’ll automatically get the recording emailed to you to watch anytime. July 9 at 12 noon EDT.
My TikTok on ‘Professional Doubters’
@aiover50 Should we all become “Professional Doubters” working with AI? A senior software engineer gave himself that title and he was right that now... See more
Worth Knowing
Voice clones are now good enough to fool the experts
Fraud tied to AI hit $442 billion last year and the part that is new is the quality. Forensic specialists say they can no longer reliably tell a cloned voice from a real one and a clone takes only a few seconds of audio to build. The "your grandkid is in trouble, wire money now" call is becoming a real concern.
Pick a family safe word this weekend, a phrase only your people know as a way to determine what is real. Read more
The people writing the AI rulebook met at the G7
President Trump and the CEOs of the major AI companies sat down at the G7 to push for a shared set of global standards. The rules that decide what your tools are allowed to do with your data are being drafted right now despite the software already being on your phones and desktops. Read more
The government just treated an AI model like a defense asset
A U.S. export-control order pulled Anthropic's top Fable 5 and Mythos 5 models offline for foreign nationals, and right after, for all users. When a chatbot gets handled like restricted military tech, you can see how high the stakes around these tools have climbed.
It's a useful reminder that the AI you use casually is now national infrastructure. Read more
52,000 Americans got asked what they really think of AI
Anthropic surveyed almost 52,000 people for its first Public Record report, mapping where people are hopeful, where they're worried and how that splits by age. The over-50 skepticism shows up as a measured position in the data. This echoes what I hear from experienced folks our age. Read more