You are not behind.
Protect your judgment.

What actually changed

• A year-long study of clinicians found that heavy reliance on an AI assistant could erode skill and confidence without an obvious warning at the time.
• The researchers described these as asymptomatic harms, which may only become visible when the person needs the underlying skill again.
• Pew reported on June 17 that the gender gap in AI use had effectively closed after men led by eleven percentage points two years earlier.
• Men were more likely to use AI daily, at 27% compared with 20% for women. Women were more likely to expect a negative personal impact, at 33% compared with 17% for men.
• Only 16% of Americans in the survey expected AI to be good for society.

Think about this

• Where are you giving AI the busywork, and where are you giving it the productive struggle that keeps your own skill alive?
• Which part of your work needs your reasoning even when AI can produce an answer faster?
• What evidence would tell you that help has started turning into dependence?

My take

Twenty years in healthcare taught me that expertise is built in the struggle. The hard part is often the part that makes you good.

I do not read the clinicians study as an argument against AI. I read it as a boundary: let AI take the toil, but keep the thinking that protects your edge.

And I do not read women's caution in the Pew data as lag. Discernment can be the thing that keeps the work yours.

What actually changed

• Kit, beehiiv, Kajabi, and Descript introduced connections that let an AI assistant draft, organize, or retrieve work from those platforms.
• Affinity added natural-language desktop instructions through Claude for tasks such as resizing, rebranding, and batch edits.
• Connector and MCP were two labels readers were seeing for the same basic idea: giving an assistant authorized access to a tool.
• The connections were opt-in. The issue recommended starting with one low-stakes task inside one platform already used regularly.

Try this: Connect one platform you already live in

• Choose the tool you opened most often that week.
• Turn on its official connection and review the access it requests.
• Ask one real question or give it one low-stakes task, such as drafting an email sequence or comparing recent click rates.
• Stop after the first test. Decide whether the connection removed work before adding another one.

My take

The win was not wiring up all five. It was choosing the one tool you already lived in and completing one useful task without exporting another file.

Fewer tools, better connected, can reduce more work than another new subscription.

What actually changed

• OpenAI said the free ChatGPT model reached the performance of its advanced models on difficult health questions.
• The system improved at flagging situations that might need urgent care, asking for context, and stating when it was uncertain.
• OpenAI reported a 71% reduction in flagged factual errors over two months and said more than 260 physicians contributed to the work.

Try this: Use it as a first read

• Ask for plain-language explanations of terms rather than a diagnosis.
• Provide relevant context and ask what information is missing.
• Ask the system to separate what it knows from what it is uncertain about.
• Bring the summary and questions to a qualified professional instead of treating the answer as the final word.

My take

A more accurate first-pass health explainer can be genuinely useful, especially when an appointment is weeks away.

Use it to walk into the room more informed, not to talk yourself out of going. First read, not final word.

What actually changed

• Notion's design lead described restrained AI as something that appears in context and gets out of the way.
• The interview treated feature overload as a design problem rather than a user education problem.
• The practical filter was simple: keep the feature you would genuinely miss and quiet the rest.

Try this: Run the quiet-tool audit

• Open the tool that has become noticeably louder.
• Turn off one AI suggestion, badge, alert, or notification you never asked for.
• Keep the one AI feature in that tool that consistently helps you finish work.
• Notice whether the tool feels easier to enter the next time you need it.

My take

Calm, contextual, and restrained is the pattern under the work I teach.

You do not have to keep up with every feature. Pick the few that matter, quiet the rest, and make the tool earn the attention it asks for.

What was coming

• beehiiv had scheduled its Summer Release Event for July 16.
• Skool was expected to have a summer release, but no date had been announced.
• Creator platforms were adding more AI capability, but readers did not need to adopt on the platform's schedule.

Think about this

• What is your settle window before you test a newly announced feature?
• What would the feature need to replace, shorten, or simplify to earn a place in your stack?

My take

Let the announcement arrive. Let the first round of problems surface. Then decide whether the feature changes work you already need to do.