Every few weeks something about AI goes off like a fire alarm. A report leaks, a headline runs, a video racks up two million views, and everyone you know spends a day in a low-grade panic.

Then the details come out, and the story is almost always smaller and stranger than the headline. But by then the audience has moved on to the next alarm, and nobody circles back to say what actually happened.

This magazine circles back. Once a month, one careful read of what was claimed, what was true, and what it means for the ordinary person using these tools to do their work.

What's coming

• OctoberThe Hugging Face Incident: No Need to Panic. Yet.
• NovemberMAX_TOKENS. Relax. Don't do it.
• DecemberShaping Prompt Output — A Guide to Better AI Code Assist
• JanuaryWhat AI Needs Is Consumers Union

Calm first. Then get to work.

Who's writing this

I've spent my career around things that can actually kill people. I did safety review for space vehicle design at NASA, including work in a 5,000-volt DC lab, where one careless moment doesn't give you a second one.

Then I moved into risk management at Enron. I met and talked with Ken Lay and Jeff Skilling. I ran the numbers, decided the risk was real, and left for a lower-paying job nine months before the collapse.

So I'm not telling you AI is harmless. I'm telling you that real risk is assessed quietly, by people reading carefully — and that almost nothing you're being shown was assembled to inform you. It was assembled to hold you.