Day 965 - When the System Can't Save You, Save Yourself
Last week’s article proved that the AI displacement trap is a collective action problem too deep for any firm to escape. This week: what that means for the rest of us — and why the most useful responses aren’t the obvious ones.
Last week, we established something uncomfortable: the companies automating jobs away are acting rationally, the system has no self-correcting mechanism, and the one policy instrument that could work requires global coordination that has never been achieved for anything.
If you found that bleak, you were reading it correctly.
But the trap operates at the level of the whole system. That is not the same as saying every individual outcome is equally determined. The aggregate is heading somewhere most of us would rather it didn’t. The question is where you are standing when it arrives.
Assets Are a Hedge, Not an Exit
The instinctive response to economic disruption is to accumulate assets. Buy property. Hold equities. Build a buffer. This is not wrong — but it is incomplete.
The demand destruction the paper describes doesn’t stop at wages. It runs through asset values too. A world where knowledge workers have lost meaningful income is a world where consumer spending has contracted, corporate revenues have compressed, and the equity portfolios tracking those revenues have repriced accordingly.
Assets help at the margin. But the more durable question is income resilience — not how much you have, but how exposed your income stream is to the dynamics the paper describes.
Where the Frontier Moves Slowest
The paper’s logic tells you which income streams are most at risk. The automation distortion is worst in fragmented, competitive markets with measurable outputs: customer support, back-office operations, entry-level software, content production at scale. These are sectors where AI deploys cheaply and output is easy to verify.
Roles harder to automate profitably share a different profile — embedded in complex human systems, where integration costs are high, output is difficult to measure, and being wrong is expensive in ways that matter to the people paying. They require not just task execution but the ability to navigate the situation around it.
The frontier will reach these roles. It is not there yet. And “not yet” is what you are working with.
The Geography of Cost
There is a response to income compression that rarely appears in serious discussions because it sounds too simple: spend less. Not as austerity — as a structural choice about how much income you actually need, and whether your current cost of living is the most efficient way to maintain what you value.
As recently as last week, my wife and I looked after two golden retrievers on well-to-do Waiheke Island for a week while their owners were away. The dogs needed to stay somewhere familiar, which meant we had to stay at the owners’ holiday house. We love the dogs. We said yes.
What followed was one of the better working weeks either of us can remember. Coastal walks every morning, a beautiful property, and a full working week without a single day of leave. Productivity, if anything, was higher than in the office. The owners got reliable, affectionate care. Nobody paid anybody anything. Everybody won.
This is cost-of-living arbitrage in its most human form. The conditions that made it possible — remote-capable work, a relationship of trust, a willingness to say yes — are conditions that can be cultivated deliberately. If income compresses, the people best placed to absorb it are not necessarily those who earn the most. They are those whose cost structure gives them the most flexibility, who have reduced their dependence on income-intensive living without reducing what they actually value. That is a form of resilience no market crash can reprice.
The Audience of One Thousand
The paper shows that the automation trap is most destructive in fragmented competitive markets, where each actor externalises the cost of their choices onto the whole. Employment sits inside that structure. An employer automating your role is not malicious. They are responding rationally to a competitive environment that punishes restraint.
A direct exchange of value between a writer and a paying audience sits outside that structure. The writer internalises both the value created and the relationship that sustains it. There is no competitor undercutting the relationship with cheaper AI, because the relationship is not a commodity. You are not selling approximable output, but a specific perspective, earned over time, delivered with accountability.
This does not make audience-based work easy or universally accessible. But understanding why it is structurally resilient, not just that people say it is, changes how seriously you invest in getting there.
The Coordination That Actually Works
The paper’s policy solution requires global coordination that has never held together for anything of comparable complexity. But coordination doesn’t only happen at that scale.
The firm-level trap exists because each actor is too small to internalise the damage they cause, and too exposed to defection to restrain themselves. A community of people with complementary skills, shared resources, and genuine mutual obligation faces a different structure: small enough that each member’s contribution is visible, trusting enough that defection carries real cost.
Mutual aid societies, guild structures, and cooperative models all emerged in periods of economic disruption where individual exposure was high and system-level protection was absent. The knowledge-work version is still being invented. But the logic is as old as the problem.
The people who navigate the next few years best are unlikely to be isolated individuals who correctly predicted the endpoint. They are more likely to be people embedded in relationships of genuine mutual value — professionally, geographically, personally — who have something to offer and people to offer it to.
The system-level trap is real. Individuals cannot escape the aggregate outcome by clever positioning. But the aggregate outcome will arrive unevenly, with significant variation in how different people experience it. That variation is not random. It correlates with the choices above — made now, while the window is still open.
Understanding why is the beginning. The rest is execution.