Surviving the AI Squeeze

Today I got an email from a friend outside tech whose work involves turning spoken notes into written records. Most of the clients they supported were recently moved to an AI tool to save money. A handful of people chose to keep working with my friend anyway, which helped, and they scrambled to pick up harder work that automated tools still struggle with, so now they are working much longer hours for a fraction of what they used to make.

Reading that email hit me hard, partly because of the week I had just been having.

A few days earlier, I was texting with two close friends in tech about pushing our AI ambition way higher after I spent ten days building a real software system with agents running around the clock. Yesterday, my dad sent me a podcast about navigating a rough job market, and I texted back feeling pretty grateful for my own footing after nearly fifteen years at the same company. And today, Anthropic went public on the Nasdaq, setting off a massive Hacker News thread full of software engineers wrestling with whether their own careers are about to change for good.

Seeing those moments land in the same week made me stop and rethink how I have been talking about AI and work.

Looking back at what I wrote a year ago

In June 2025, after a heavy conversation with a newer engineer who worried they had picked the wrong career, I wrote The Uncertain Future of Coding Careers and Why I’m Still Hopeful. A few months later, after layoffs hit close to home again, I wrote Building Career Security in an Uncertain World.

I meant everything I wrote in those posts. I believe human creativity and judgment matter, and I have seen firsthand how much more a single person can build when they learn to work well with agents. I also meant what I said about job security being fragile and career security being something you have to practice on purpose.

Writing from inside software engineering gives you a comfortable vantage point, though. If you already have decent pay, autonomy, supportive leadership, and access to good tools, it is easy to tell people to become masters of context and shepherd agents. That advice sounds a lot thinner on a Thursday afternoon when a friend outside tech just watched most of their income disappear because an organization wanted to cut costs through automation.

Where the squeeze is actually happening

In the Hacker News thread on Anthropic’s IPO today, people kept arguing over whether software engineers are paid to solve problems or paid to write code. One commenter said something that applies way beyond tech. Plenty of people in every field are paid to reliably do the thing in front of them, whether that is connecting two databases, transcribing notes, processing claims, or reconciling spreadsheets. When a business sees a way to buy that output cheaper with AI, it often chooses cheaper over better.

Even inside companies that are enthusiastic about AI, the transition is messy. Two friends at other tech companies told me this week that management expectations are speeding up so much that teams barely have breathing room to learn a different way of working. People try an AI tool once, get a mediocre first draft, and drop it because they have a deadline staring at them. So you end up with a widening gap between the handful of people who have space to experiment and everyone else who is just trying to keep their head above water.

I want to be honest about my own side of this too. In a conversation with a colleague last year, I caught myself admitting that I have been operating in constant-reinvention mode for the last few years partly because I do not want to lose my job. Even from a strong seat, the ground feels like it is moving.

Three ways the next few years could go

Nobody knows how fast this plays out, and I distrust anyone who claims they do. When I try to think honestly about what my daughters, my friends, and I are walking into, I keep coming back to three rough possibilities.

The worst case (besides we all die)

Companies treat AI mainly as a way to cut payroll and squeeze margins. Any role built around repetitive, predictable work gets compressed, so fewer people are hired and the people who stay end up working longer hours for less pay. A lot of human work turns into stressful cleanup, handling the messy edge cases that automated tools miss, while entry-level jobs disappear and make it harder for younger workers to ever build real experience.

The middle case

This is the one I think is most likely, and it is already starting. We get a split economy. Routine, repeatable work gets cheap very quickly, and new kinds of work do emerge over time, because people never run out of problems they want solved. The transition is uneven and rough, though. People with autonomy, strong relationships, and trust from the people they serve end up doing very well, while people in rigid roles or competing mostly on price get squeezed hard. A few years in, organizations realize they starved the pipeline of people who actually understand how their systems and businesses work.

The best case

Cheap automation lowers the barrier to building small, durable businesses and solving specific problems that never used to justify the cost. Instead of only big companies capturing the gains, ordinary people and small teams use AI to handle the drudgery underneath their work while spending more of their energy on craft, relationships, physical work, and judgment. Work feels more human, because the parts people actually pay a person for are trust, taste, and accountability.

How to give yourself a better shot, whatever you do

When my dad texted me that podcast yesterday, he pointed out one line near the end. When the host asked how people should think in a tough job market, the guest gave two answers, build AI skills and work toward a side venture using your skills.

I texted him back with the version I hope to teach my daughters. Find a problem people actually have that they are willing to pay someone to solve, make sure it overlaps with something you genuinely care about, and then incrementally use AI to deliver that value better over time.

Regardless of what field you are in, there are a few practical moves that push you closer to the best case and protect you if we land in the middle or worst one.

Move closer to trust and accountability

Think about why a handful of people chose to keep working with my friend even when they were offered an AI alternative. They trusted the accuracy, and they knew my friend understood how they worked. And the harder work my friend had to scramble for afterward only existed because real life is full of messy edge cases where generic tools fall short.

In any career, the fragile spot is interchangeable work where someone can swap you out for a cheaper vendor and barely notice. The steadier ground is where a person or a customer specifically trusts your judgment, your relationship, and your willingness to stand behind the outcome. Look at your current work and ask which parts people buy from you because they trust you, and try to grow those parts.

Use AI to cross a boundary

If you only use AI to do your existing task 30% faster, your employer often pockets the savings, or eventually decides a smaller team can cover the same load. The bigger opportunity is using AI to step into adjacent work you could not do alone before.

A contractor or tradesperson can use AI to handle quoting, scheduling, and customer follow-up that used to eat their evenings. A teacher or nurse can build small tools that take hours of paperwork off their plate. A bookkeeper can move toward helping small businesses clean up messy operations and cash-flow decisions instead of only entering receipts. An engineer can step closer to customers and product decisions instead of waiting for tickets. Crossing one boundary changes what you can offer.

Push yourself while you still have runway

It is really easy to settle into a comfortable routine with AI, asking a chatbot a few questions a day and feeling like you are keeping up. I noticed I had fallen into that exact comfort zone for a couple of months this fall before pushing myself to try much bigger things recently.

Practicing on real problems while your job is still steady is a lot easier than scrambling after a layoff or a sudden drop in clients. Pick one annoying problem in your work or your daily life every week and push the tools harder than you think they can handle. Even when an experiment fails, you learn where the real limits are, and that knowledge is valuable on its own.

Build a small paid problem on the side

I see too many people approach side projects from one of two extremes. Either they build something that scratches their own itch with no real path to steady income, or they chase money in something that brings them no joy and burn out fast.

Even a small side effort that solves a real problem for real paying customers changes how you see your career. You learn how to talk to buyers, price your work, and deliver value directly instead of relying completely on one employer’s budget spreadsheet. You may never need it to replace your day job, and having that muscle built ahead of time still changes your options.

Protect your ability to actually understand the work

Earlier this year I wrote The Work Can Look Done Before the Thinking Is Done, and I think that warning applies to every profession right now. Agents are great at producing polished drafts, clean code, and confident summaries before anyone has really checked whether the underlying work makes sense.

If you let the tool do all of your thinking and turn yourself into a rubber stamp, you become easy to replace by anyone else who can type a prompt. Keep your hands close enough to the real work that you can spot when the machine is confidently wrong. As more people hand over their judgment to automated tools, the people who still truly understand how things work will become rarer and more valuable.

Keep investing in real people and local reputation

When resumes, cover letters, and online content can all be generated in seconds, it gets much harder to stand out through volume. Real relationships and local reputation do the opposite. People still hire, refer, and stick with humans they actually know and trust.

Show up in your community, help people when you do not need anything back, and share what you are learning out loud when you can. Most of the best turns in my own career came from genuine friendships that started years before anyone offered me a job.

What this asks of people who already have room to breathe

If you are reading this from a fortunate spot like mine, with a steady paycheck, supportive managers, and freedom to experiment, I think we owe people more than pep talks.

Give the people around you actual time and space to learn these tools instead of only turning up the delivery pressure. Keep hiring and mentoring people earlier in their careers, even when it feels faster in the short term to let an agent do the small tasks yourself. And be honest about who is taking the hit right now. Pretending everyone can just learn a few prompts and land on their feet ignores real people who are working twice as hard for half the pay.

I do not know which of those three futures we are going to get. My guess is we will live through a messy mix of all three at the same time, depending on where you sit and what kind of work you do. I am still figuring out how to talk about all of this with my daughters as they grow up into it. What I know today is that hoping for the best case is not enough on its own, and looking away from the people already getting squeezed does not help anyone either.

Further thinking

  • · Justin Searls
    Surviving the AI Squeeze

    Justin responds directly to the widening gap inside tech companies and pushes harder on why traditional fixes fall short. He argues that corporate training cannot solve an identity and mindset shift, that single-player agentic work makes mentoring harder, and that executives, managers, and developers all have to act like the job has already changed.