An AI-first memo is a hiring freeze

The most important sentence in Shopify's AI memo is not about AI. It says a team must show why AI cannot do the work before it asks for more people. Read past the vocabulary of this spring's AI-first memos and each one is a hiring freeze that also grades you on chatbot usage.

I was a new-grad hire eleven months ago. I finished my master's in Chicago in May 2024 and a month later started as a software engineer in a university IT department. Score me on Shopify's rubric and I would do fine, because a chat model sits in a window beside my editor most of the day. That is exactly why I distrust the rubric. I am also the kind of hire these memos make harder to approve.

Tobi Lütke posted the memo on X on April 7, after it had started to leak. Its headline is "Reflexive AI usage is now a baseline expectation." Parts of it are right. An engineer who has not tried these tools by 2025 is being stubborn. But two rules sit under the headline. One makes AI usage part of performance and peer reviews. The other is the headcount rule.

The first rule is Goodhart's law with a logo. Once a number becomes a target, people hit the number, and the number stops telling you anything. Score AI usage in reviews and you will get engineers who route every ticket through a chat window where the usage shows up, whether or not it helped. You will not learn whether the work got better, because you never measured the work. It is like grading a writer by how often she opens the thesaurus.

In January, back home, the chairman of L&T suggested a 90-hour week, Sundays included, and asked, "How long can you stare at your wife?" L&T's HR later said he had not mandated 90-hour weeks and that the remarks were casual. Take them at their word. A casual remark still shows what a chairman counts, and he counts hours at a desk the way these memos count prompts, because a number is easier to read than a decision about what the team should build.

The second rule is the real memo.

A manager who wants a person now has to prove a negative, that no tool can do this work, and nobody can prove that in a year when a new model ships every few weeks. The request dies, and nobody has to sign anything with the word "freeze" in it. I call this headcount theater.

Salesforce did it first. Marc Benioff said in December, and again on the February earnings call, that Salesforce would add no software engineers in 2025 because AI had raised engineering productivity "by more than 30%." Measured how? Salesforce has not published a method. He also planned to hire 1,000 to 2,000 more salespeople. By his own number, engineering got more than 30% more productive, and the new seats went to sales. On April 28, Luis von Ahn declared Duolingo "AI-first," with contractors phased out for work AI can do and AI use weighed in hiring and performance reviews. Users answered by deleting their long streaks in public.

Then there are two dates I keep putting side by side. On April 29, Satya Nadella said that 20 to 30% of the code in Microsoft's repositories was "written by software," a range ten points wide. Last Tuesday, Microsoft cut about 6,000 jobs, about 3% of the company, and said it was cutting layers of management. In the WARN filings for Washington state, the largest group was software engineering, 817 of about 2,000 local cuts. Microsoft did not say AI caused any of it. It didn't have to.

Junior work fails the headcount rule first. A first-year ticket, small and scoped and a bit dull, is exactly what a chatbot demo handles well. A manager who asks for a staff engineer can still argue for judgment. A manager who asks for a new grad loses the argument to one screen recording. But nobody hires a junior for the tickets. The first year is how a company makes its seniors.

The juniors were in trouble before any memo. The New York Fed's table puts unemployment for recent computer science graduates at 6.1%. That data is from 2023, before most coding agents existed, so I won't pin it on a chatbot. The memos hand every manager a sentence to quote when the next new-grad request comes in.

A company can freeze hiring for honest reasons. Money gets tight, or a company hired too fast in 2021, and then someone signs the freeze and answers for it. Headcount theater is the same freeze minus the courage, and the graduate who never gets the offer hears that a machine decided.

If you mean freeze, write freeze.