In this first of three articles covering the biggest week in AI this year:

Reading Time: 12-15 minutes | Listening Time: 10-12 minutes if read aloud

Three major things happened last week. Each one deserves its own article. So that is what we are doing.

This edition covers Anthropic's landmark study on AI and jobs. In Edition 335, I am going deeper on what cognitive partnership actually means now that the infrastructure for it exists. And in Edition 335, we will look at OpenAI's GPT-5.4 release and what a million token context window means for knowledge workers.

But we are starting here. With careers. With jobs. With how not to get replaced.

And I need to tell you something. The timing of this study could not be more perfect.

In Edition 333, I told you about the morning I spent four hours in Cowork mapping three years of my own work. I told you about the 25 tools I had open, none of which knew who I was. I told you about Anthropic releasing the Memory Tool and Import Memory. And I launched the Cognitive Partner Membership because I realized the clean path I built that morning was something I could hand to other people.

One day later, Anthropic published the data that explains why all of that matters.

On March 5th, Anthropic's research team published a study called "Labor Market Impacts of AI: A New Measure and Early Evidence."

This is not marketing. This is not a blog post with opinions.

This is real economic research. Bureau of Labor Statistics data. Their own platform usage data. Established academic frameworks. Peer-reviewed methodology.

They created something called "observed exposure." It measures which jobs are actually seeing AI used in professional settings right now. Not what AI could theoretically do. What it is actually doing.

Here is what they found.

AI is far from reaching its theoretical capability. The gap between what AI could do and what people are actually using it for is enormous. Even in the most exposed category, Computer and Math occupations, Claude currently covers just 33% of all tasks. The theoretical capability is 94%.

That gap is everything. I will come back to it.

Computer programmers are the most exposed occupation at 75% task coverage. Customer service representatives are next. Then data entry workers at 67%. Financial analysts are in the top ten.

30% of all workers have zero AI exposure. Their tasks showed up too infrequently to even register. This includes cooks, motorcycle mechanics, lifeguards, bartenders, and dressing room attendants. If your job involves physical presence and hands-on work, AI is not touching it yet.

No systematic increase in unemployment for highly exposed workers since late 2022. Read that again. Despite everything you have heard, the data does not show a wave of job losses in AI-exposed occupations.

But hiring of younger workers aged 22 to 25 has slowed in the most exposed occupations. This is the early warning signal. Not layoffs. Slower hiring. A 14% drop in the job finding rate for young workers in exposed fields compared to 2022.

Workers most at risk tend to be older, female, more educated, and higher paid. People with graduate degrees are almost four times more represented in the most exposed group compared to the unexposed group. The exposed group earns 47% more on average.

This is not about factory workers.

This is not about truck drivers.

The people with the most education and the highest salaries are in the most exposed positions.

That flips a lot of assumptions on their heads.

If you went to college, got an advanced degree, built a career in knowledge work, and thought that made you safe from automation... this data says think again.

But here is the thing.

It also says you have time.

Here is the part most people will miss.

Anthropic's own data shows that actual AI usage is still a fraction of what is theoretically possible. The gap between capability and adoption is massive.

That gap is your opportunity.

And now think about what I described in Edition 333. Twenty-five tools. None of them knew who I was. Every morning, re-explaining myself to AI that forgot me overnight. That is the gap in action. The tools are capable. But most people are not using them in any kind of integrated, intentional way.

That is exactly why I built the Cognitive Partner Membership. Because closing that gap is not about having more tools. It is about having the right setup. The Single Source of Truth from Edition 329. The cognitive profile that travels with you. The system where you open one tool and it already knows you.

The people who close that gap between AI capability and actual usage? Those are the people who stay relevant.

This is what I was getting at in Edition 330 when I talked about the world of abundance that is coming. The question I asked then was philosophical: what do humans do when the problems are solved? Today's data makes it practical. The problems are not all solved yet. But the path is clear. And the time to position yourself is now.

Nobody is getting replaced by AI.

People will get replaced by other people who know how to use AI.

That is not the same thing. And it is an important distinction.

I am not building AI tools to replace people. I never have been. Everything at LM Lab AI is built around the idea that these tools should help you stay relevant, find more opportunities, and do better work. Not take your job.

The Anthropic study confirms what I wrote about in Edition 331 with real numbers.

Two groups are struggling right now for completely opposite reasons.

The first group is young. 22 to 25. Just entering the workforce. AI literate but lacking domain expertise. And the data shows they are finding it harder to get hired into the jobs that AI is touching. That 14% drop in the job finding rate is the number behind what I described in Edition 331.

The second group is experienced. Deep knowledge. Decades in their field. But many have not touched these tools. And their occupations are the most exposed.

Both groups have exactly what the other one needs.

In Edition 331, I called this the "meet in the middle" moment. Today we have the numbers behind it. The experienced workers need AI fluency. The young workers need domain expertise. The data shows both sides are struggling in isolation. The path forward is together.

If you are an experienced professional in your 40s or 50s, you are actually in the best position right now. You have something AI cannot replicate. Years of pattern recognition. Institutional knowledge. Relationships. Judgment that only comes from doing the work.

Add AI fluency on top of that and you become a force multiplier.

That is exactly what I did in the Cowork session I described in Edition 333. I brought 15 years of recreation business experience and three years of AI research to a four-hour conversation with a tool I had never used before. And I walked out with a complete business strategy. The domain expertise made the AI useful. The AI made the domain expertise scalable.

The people who figure this out in the next two or three years are going to be incredibly valuable.

If you are a younger worker, your advantage is that you are more naturally comfortable with these tools. But you need to pair that with real domain knowledge. Find mentors. Learn from the people who have been in the trenches. Their expertise plus your adaptability is a powerful combination.

The people who will struggle are in neither camp. The experienced professionals who refuse to learn. And the young workers who rely only on AI without building real expertise underneath.

Both extremes lose. The middle wins.

Now here is where it gets personal.

Everything I just described requires a specific cognitive skill. The ability to adapt how you think. Learn new tools quickly. See connections between different domains. Pivot when things change.

That is cognitive flexibility. And if you are neurodivergent, especially if you are dyslexic, you have been training that muscle your entire life.

I wrote about this in Edition 323 when I talked about going against the grain my whole life. The same lateral thinking that made school difficult is now the exact skill that AI cannot replicate.

Think about it.

Every time you found a workaround because the "standard" way did not work for your brain, you were practicing cognitive flexibility. Every time you figured out how to get from Point A to Point B using a completely different route than everyone else, you were building the exact skill set that this new economy demands.

The Anthropic study shows that the most exposed jobs are the ones built around sequential, repeatable cognitive tasks. Data entry. Customer service scripts. Routine programming. These are the tasks that AI handles well because they follow predictable patterns.

You know what AI is terrible at?

The things dyslexic thinkers tend to be great at. Lateral connections. Seeing the big picture when everyone else is stuck in the details. Pattern recognition across completely unrelated domains. Strategic framing. Reading the room. Understanding context that does not exist in a dataset.

This is exactly what the Cognitive Balance Model is about. I replaced the 10-80-10 Rule with this framework because it better captures what is actually happening. You initiate the work by seeing what needs to happen. That is Human Initiation. AI handles the execution, the formatting, the sequential processing that creates friction for our brains. That is AI Expansion. Then you come back in for the judgment calls, the pattern matching, the strategic decisions. That is Human Integration.

The tools are not here to replace you. They are here to handle the parts that were never your strength anyway, so you can focus on the parts where you are irreplaceable.

In Edition 335, I am going deeper on how cognitive partnership works now that we have the Memory Tool and Import Memory. The Cognitive Balance Model gets significantly more powerful when your AI partner actually remembers how your brain works.

Here is what I would tell anyone reading this today.

Stop thinking about AI as a threat. Start thinking about it as a tool you need to learn. Like you learned email. Like you learned smartphones. Like you learned whatever the last big technology shift was.

Identify what you bring that AI cannot. Your relationships. Your judgment. Your domain expertise. Your ability to read between the lines. Lean into those things.

Use AI to handle the cognitive load. Delegate the repetitive, sequential, energy-draining tasks. Keep the creative, strategic, high-judgment work for yourself. This is the Cognitive Balance Model in practice.

Set up your cognitive partner properly from the start. If you have not read Edition 333, go back and read it. The founding member offer for the Cognitive Partner Membership is still open. Fifty spots at $19/month locked forever. This is the clean path I wish I had when I started. You do not have to build the mess first.

Recognize this is a lifelong path now. The tools will keep changing. The models will keep improving. In Edition 336, I will break down GPT-5.4 and what the pace of these releases means. Your job is not to master one tool. Your job is to stay adaptable. And that is something neurodivergent minds have always been good at.

Share what you learn. Whether it is with a colleague, a younger worker, or someone in your field who has not started yet. The "meet in the middle" from Edition 331 only works if people are willing to teach and willing to learn.

Anthropic's own research shows the sky is not falling. Not yet.

The gap between what AI can do and what is actually happening in the real economy is still enormous.

That gap is where you position yourself.

In Edition 333, I showed you what it looks like when you close that gap for yourself. One morning. One tool. Three years of work, finally connected.

Today, the data says the same thing from the other direction. The opportunity is real. The window is open. And the people who move now will be the ones who thrive.

Not by fighting against AI. Not by pretending it does not exist. And not by assuming it will do everything for you.

By learning to work with it. By bringing your unique cognitive strengths to the table. By staying flexible.

If you have spent your whole life figuring out how to think differently, congratulations. That is exactly the skill the future demands.

Anthropic: "Labor Market Impacts of AI: A New Measure and Early Evidence" (March 5, 2026) anthropic.com/research/labor-market-impacts

This is Part 1 of 3 from the biggest week in AI this year.

Previously: Edition 333 covered the Cowork session, Import Memory, the Memory Tool, and the launch of the Cognitive Partner Membership. Edition 331 explored the generational crossroads. Edition 330 asked what problem solvers do when AI solves everything. Edition 329 introduced Building Your Second Brain and the Single Source of Truth.

Next: Edition 334 goes deeper on cognitive partnership and what AI memory means for neurodivergent thinkers. Edition 336 breaks down GPT-5.4 and the acceleration of AI.

Dictated, not typed. Obviously.

📊 The Study: Anthropic published real economic research on AI and jobs using Bureau of Labor Statistics data. This is not opinions. This is data.

🖥️ Most Exposed: Computer programmers (75%), customer service reps, data entry workers, financial analysts. Knowledge workers, not factory workers.

📉 No Mass Layoffs Yet: No systematic increase in unemployment for exposed workers since 2022. But hiring of 22-25 year olds has slowed 14% in exposed fields.

🎓 Surprising Finding: The most exposed workers are more educated and higher paid. Graduate degrees are 4x more represented in the exposed group.

🪟 The Gap Is Everything: AI usage is still a tiny fraction of what is theoretically possible. The gap between capability and adoption is your opportunity. This is the same gap I showed you in Edition 333 with 25 tools that did not know who I was.

🤝 Meet in the Middle: Experienced workers need AI fluency. Young workers need domain expertise. Both groups have what the other needs. (See Edition 331 for the full framework. See Edition 333 for what closing the gap actually looks like.)

🧠 Neurodivergent Edge: The most exposed jobs are sequential and repeatable. The least automatable skills are lateral thinking, pattern recognition, and strategic framing. Sound familiar? (See Edition 323.)

⚖️ Cognitive Balance Model: Human Initiation + AI Expansion + Human Integration. Let AI handle what drains you. Focus on what makes you irreplaceable.

🚀 Founding Members: The Cognitive Partner Membership from Edition 333 is how you close the gap. 50 spots at $19/month locked forever.

🎯 Bottom Line: Nobody is getting replaced by AI. People will get replaced by other people who know how to use AI. Learn the tools. Stay flexible. That is the path.

➡️ Next: Edition 336 on cognitive partnership. Edition 336 on GPT-5.4.