​

This past week, four companies announced layoffs. Same industry pressure. Same AI narrative in every press release.

Completely different behavior.

One company released a TikTok of a boss kicking a cup of coffee into an employee's face. Seven days later, they fired 22% of their staff. Another company locked 1,600 employees out of their laptops at 7am without sending a single email. Workers found out they were fired because their computers stopped working.

A third company was losing money, and still gave every departing employee full base pay through the end of the year, continued healthcare, and equity vesting past their departure date.

Same economic pressure. Radically different character.

And here is the thing that keeps getting buried in the headlines: none of these companies cut people because AI replaced their jobs. They cut people to fund AI infrastructure. Data centers. Compute. Capex. The people weren't replaced by the technology. They were traded for the budget to build it.

That is a different problem. And it points to a different answer.

What the layoff pattern actually reveals

The companies behaving badly share one thing. They are treating AI like a headcount swap. Fire people, deploy tools, maintain output. It is a short-term math problem dressed up as a transformation strategy.

The companies doing it differently are thinking about something else entirely. They are asking what work actually requires a human in the loop, where judgment lives that a model cannot replicate, and how to build systems around the people they keep rather than just reducing the count.

That distinction matters for you, regardless of where you work.

Because your company is watching the same headlines. Your leadership is having the same budget conversations. And the question of who stays and who gets caught in the next restructure is not going to be answered by who uses AI the most.

It is going to be answered by who understands how to build with it.

The shift most professionals are missing

There is a version of using AI that looks productive and produces almost nothing durable. You send a prompt. You get an output. You send another prompt. You get another output. Each one is fine. None of them connect. You are not building anything. You are just moving faster through a task list that resets every morning.

The professionals pulling ahead are not prompting more. They are doing something different. They are designing the loop first.

They start with the output they need.
Then they work backward to the input.
Then they build the connection between them in a way that runs without them having to restart it from scratch each time.

That is not a technical skill.
It's a thinking skill.

And it is the one that makes the difference between someone AI makes more productive and someone AI makes more replaceable.

The practical move this week: design the loop, not the prompt

Most people try to improve their AI results by writing better prompts. That is the wrong level to work at.

Here is the move that actually compounds:

Pick one recurring output from your work.
Not a task, an output.

A status update that goes to the same people every week.
A summary that always pulls from the same kind of source.
A first draft that always follows the same structure.

Name what the finished thing looks like before you touch any tool.

Every recurring output has a trigger.

A meeting ends. A report lands in your inbox. A deadline hits.

Write down the trigger.

You now have the beginning and the end of your loop.
Everything in the middle is the system.

Before you build anything, run the loop by hand using AI.

Give it the trigger input.
Describe the output you need.

See if it holds.

If the manual version breaks, the automated version will break worse. Fix it here first.

Once the manual version works consistently, look at each step and ask one question:

Does this step require my judgment, or just my presence?

The steps that only need your presence are the ones you automate. The steps that need your judgment are the ones you stay in.

This is the step most people skip.

They automate everything they can and then wonder why the output feels hollow.

The judgment layer is where your value lives.
​
Protect it.
Name it explicitly.
Make sure it is visible to the people who need to see it.

The operators who are building durable positions right now are not the ones who automated the most.

They are the ones who figured out where their judgment actually lives and built systems that protect it.

I know we went over a lot today, but man I hope it was helpful and insightful!

(If it sucked, pleased respond and let me know - I genuinely love all the feedback you send and want to make the next issue even better)

One Quick Question, and One Ask

Question: What is one recurring output in your work right now that has a clear trigger and a clear finished form?

Hit reply and tell me - I try to respond to as many of you as I can!

One Ask: If this framing was useful, forward it to someone on your team who is trying to figure out where AI fits in their actual job.

It's a win-win: it makes you look smart for sharing and also makes them look smart for reading!

​