The Two-Line Person
The Shortest Program
Model a person's decision-making as a function. Not their full personality -- just the core loop. The part that takes in a situation and produces an action.
Some people's functions are long. Conditionals, internal conflicts, recursion, reference to external standards. The code branches. It struggles. It sometimes loops without terminating. Messy, but alive.
Other people's functions are two lines of Lisp:
(defun person (input)
(conform (read-air)))
The input is ignored. It does not matter what happens -- economic crisis, pandemic, war mobilization, policy reversal. The output is always the same: read the air -- sense what everyone else is doing -- and conform. No branching. No internal state. No struggle function.
The Debate That Misses the Point
The AI conversation keeps circling the same question: which jobs get automated?
Programmer, designer, copywriter, accountant. The lists never stop. Everyone maps their own occupation against the threat. It is understandable. It is also the wrong frame.
The deeper question is not which tasks AI can perform. It is which people are already performing like machines -- and what happens when the system that manages them gets an upgrade.
A person running (conform (read-air)) is already automated. Not by silicon, but by social architecture. The compliance is built in. The self-policing is built in. The absence of independent judgment is built in. AI does not need to replace this person. It needs to manage them more efficiently.
Conformity as Pre-Automation
We talk about AI as a force that makes people redundant. But some populations were made redundant long before AI arrived. Their decision architecture was stripped down generations ago -- by education systems that optimize for compliance, by corporate cultures that absorb individual identity, by social pressure mechanisms so distributed that nobody is responsible for them.
These populations already run on minimal code. They respond to authority cues, social consensus, and fear of exclusion. No sophisticated persuasion needed. Just shift the signal and they follow.
AI does not threaten these people in the way we usually mean. It does not take their jobs and leave them struggling. It completes the circuit. The population was already engineered for control. AI just makes the control layer smarter, faster, and cheaper to operate.
Governments and corporations currently spend enormous resources on consensus manufacturing -- media, PR, advertising, policy framing. All of that exists because even compliant populations still have friction: people get confused, messages get distorted, local variation creates noise.
LLMs eliminate that friction. Personalized messaging at scale. Real-time sentiment adjustment. Content generation that matches each individual's conformity profile. The read-air function gets a better feed.
The Branching Premium
Not everyone runs on two lines. Some people have conditionals. They check input against internal standards. They have a struggle function -- a place where competing values create genuine conflict before action. The code is longer, harder to predict, and harder to manage.
These people are expensive. Not in salary -- in controllability. They push back. They ask questions that break the script. They make decisions that do not match the signal. From a management perspective, they are bugs.
But from a survival perspective, they are the only ones who can adapt when the environment changes in ways the signal did not predict. The two-line person follows the air even when the air is wrong. The branching person has a chance of noticing.
This is the premium that matters in an AI-saturated world. Not "skills AI cannot do" -- that list gets shorter every time you check. The premium is on decision architecture that is not reducible to conformity. Internal complexity. The willingness to produce output that does not match the consensus input.
Who Gets Managed, Who Gets to Manage
AI tools are sold as empowerment. "AI for everyone." Democratized intelligence. But the actual deployment pattern tells a different story. AI that empowers individual decision-making -- coding assistants, research tools, creative aids -- flows to people who already have complex decision architectures. People who know what to build and why. People with judgment.
AI that manages and directs -- recommendation engines, content feeds, behavioral nudges, automated HR screening -- flows toward populations that are already compliant. The tool is the same. The function is opposite.
The people who were already branching get better instruments. The people who were already conforming get a better-optimized cage.
Systems optimize for what is measurable, and compliance is very measurable.
The Test
When the consensus shifts -- on a technology, a policy, a social norm -- do you check it against something internal before following? Or do you update your position to match the new air automatically?
If you have never noticed yourself doing the check, you might not have one.
The test is not whether you end up agreeing with the consensus. Often the consensus is correct. The test is whether there is a step between reading the signal and producing your response. A conditional. A moment of genuine evaluation.
That step is the entire difference. It is the line between being managed by AI and using it.
Two Lines, Forever
The two-line person has always existed. Every society has them. The ratio varies.
What AI changes is the cost of managing them. The institutions, hierarchies, and broadcast media are still there. AI just makes the whole stack run with less effort and fewer people.
For the branching person, the leverage has never been higher. For the two-line person, nothing visibly changes. The air still flows. They still read it. They still conform. The cage got an upgrade, but the cage was always invisible anyway.
How many people actually have branching logic, and how many were already running on two lines long before AI showed up?