Quick note on something I've been building, and then I'll get out of your inbox.

I've spent a lot of this year writing about AI and jobs, and I've become steadily less satisfied with how that subject gets covered — including by me. The genre runs on a striking anecdote, a scary number with no denominator, and a conclusion that was clearly written before the evidence turned up. It's very hard to tell whether things are actually accelerating or whether you just read three alarming articles in a row.

So I built a scored model instead.

It tracks eighteen signals every week — the ratio of entry-level to senior job postings, offshore services headcount, staffing volume, prime-age labour force participation, whether productivity gains are reaching wages or accruing to capital — and scores each one against written anchors that don't move. The point isn't the number. The point is that the direction becomes legible instead of anecdotal.

Here's the thing that convinced me it was worth doing. Layoffs are running near historic lows — the four-week average of new jobless claims just hit its lowest level since mid-May. Almost nobody is being fired. At the same time, the average duration of unemployment climbed to around 25 weeks from about 21 a year earlier, and prime-age participation fell six tenths of a point in a single month while the unemployment rate went down — meaning people left the labour force rather than showing up as unemployed.

Here's one more, from this week. Staffing employment is booming — hours up 9% year over year, the strongest growth since 2022, with industrial staffing up 16%. Every category is up except one. Office and clerical hours are down 5%.

That combination is invisible in the headline number. It's also exactly what displacement looks like when it runs through hiring freezes and roles that quietly stop being backfilled, rather than through layoffs.

The full methodology is published, free and unpaywalled — every indicator, every source, every scoring anchor, and every known weakness in the model. If you want to check whether the thing is any good before you take a single reading seriously, that's the point.

Next Monday I'll publish the first reading, and it's not the one I expected to publish. The model found a flaw in itself in its first scored week. I'm putting out the flaw, both scorings, and the correction — including the part where my estimate of the correction was also wrong.

Back to the regular programme in the meantime.

The president of OpenAI. The CEO of Google. The founder of LinkedIn. We talk to the people building AI. Every morning, we put what they told us in your inbox.

What they're actually worried about. What they're betting on. What's coming next. Delivered in five minutes, before your first meeting.

With one email, and only five minutes, you can stay ahead of 99% of the world.

Join 75K Readers →

Nobody became a developer to write documentation. But the docs still need to get written — PRDs, README updates, architecture decisions, onboarding guides.

Wispr Flow lets you talk through it instead. Speak naturally about what the code does, how it works, and why you built it that way. Flow formats everything into clean, professional text you can paste into Notion, Confluence, or GitHub.

Used by engineering teams at OpenAI, Vercel, and Clay. 89% of messages sent with zero edits. Works system-wide on Mac, Windows, and iPhone.

Try Wispr Flow free

Skip the sales call. Walk through Gladly's interface yourself — the AI suggestions, the unified customer view, the full conversation thread. 15 minutes, no installation, no commitment.

See it yourself