The Dependent, the Avoider and the Amplifier

Type one The Dependent

What it looks like

A person who uses AI for everything. Asks the tool the question, accepts the answer, ships the output. Looks productive. Produces a lot, very fast. Receives praise and visibility, especially in environments that measure output rather than capacity. Speaks fluently about AI tools and AI-assisted workflows. Does not, when pressed for technical depth, have it.

In the book this is Abhi: the batchmate who cuts a migration script's processing time by forty percent in a day and a half, is named in the quarterly review for "initiative", and then pauses when a senior architect asks why he chose that approach for handling race conditions. "It's a standard pattern," he says. Which is not wrong. It is also not an answer.

How dependency forms

Dependency is not a moral failing and it is not laziness. It builds itself one good outcome at a time, and the people most likely to fall into it are the ones who care about good outcomes. You have a task. You are tired, or busy, or curious. You give it to the tool. The output is polished — better than what you would have produced in the time. You ship it. Nothing breaks. Something good happens: you finish faster, the recipient is pleased, maybe you are praised. Your brain registers that worked. Next time you reach for the tool slightly more readily. Over weeks the friction becomes nothing. The tool becomes the default. Your thinking, which used to come first, starts coming second — if at all.

What makes this dangerous is that the loop is positive throughout. There is no negative signal. The thinking the tool did was invisible to begin with, so the absence of your thinking is also invisible. You cannot miss what no one sees.

The three things the Dependent loses

The ability to evaluate. If you did not think about what you wanted before the tool produced something, you cannot tell whether what it produced is good. You know it is polished. You have no internal standard to judge it against, because the criteria were never built.

The ability to defend. When someone asks you to explain, modify or extend the work, you cannot. You have the output, not the structure. People sense this within minutes of a real question, and the important work quietly gets routed to someone else.

The ability to grow. Skills compound through deliberate practice. The Dependent skips the struggle and gets the output without the climb, so the climb does not happen. Two years in, they are producing the same kind of work they produced in month one, just faster. In a world where the bar moves every six months, no growth is decline.

In the book: introduced in Chapter 4, treated in depth in Chapter 5, where Arjun finds his own miniature version of it — a Tab pressed in Cursor without reading, and three weeks of status updates polished by ChatGPT into a voice that was "a slightly upgraded version of how he might write if he were trying to sound like someone who knew what he was doing".

Type two The Avoider

What it looks like

A person who refuses to engage with AI in their work. Sometimes principled. Sometimes proud. Sometimes inertial. Often the most careful, most experienced person in the room. From outside, the Avoider looks responsible — the adult who is not falling for the hype. For a period, that reading is accurate. After that period, the same posture that read as wisdom begins to read as something else.

In the book this is Karan: a team lead who reviews pull requests line by line and told Arjun in his second week, "If you can't explain it, you don't ship it. That's not a rule. That's just engineering." When the AI directive came down, Karan said, reasonably, "We're not going to rush into tools we don't understand." Evaluate-and-adopt became evaluate-and-wait, and then simply wait. Three people on his team were let go.

The three flavours of avoidance

The Fearful Avoider believes the technology is dangerous, ethically compromised, or about to displace the things that matter. The fear may be real or overblown; either way it functions as a reason not to engage, and the longer the non-engagement continues the harder it is to tell whether the fear was ever warranted.

The Proud Avoider believes real work does not need the new tools. Real engineers write their own code; real writers write their own sentences. The pride is built on a real attachment to craft. It is also, often, the disguise that lets a person not face that the world is changing in ways they have not absorbed.

The Lazy Avoider means to learn the tools, will learn the tools, has been meaning to for several months. The not-learning is not a position; it is a postponement that becomes a position by default. This is the easiest Avoider to help, because they are not committed to their avoidance.

How avoidance forms

Avoidance, unlike dependency, does not begin from convenience. It begins almost always from a real value — quality, craft, discipline, privacy, ethics, care for the people you work with. The first instinct to slow down before adopting a powerful tool is an excellent instinct. Then something happens. The thing being protected starts being used as a reason not to engage at all. The valid first refusal becomes a reflex. The careful evaluation becomes a permanent posture. At no point does the Avoider feel they have changed. They still feel like the protector. They have calcified around the value.

What it costs — and who pays

The Avoider does not lose themselves the way the Dependent does. Their craft stays intact; the work they ship today is still competent. What they lose is relevance, slowly, in the gap between the kind of work they are still doing and the kind being done elsewhere. They do not fall off a cliff. They watch, from a slowly shrinking island, as the rest of the work evolves without them.

And avoidance is contagious. If the Avoider has any authority — manager, team lead, teacher, parent, senior figure of any kind — their team takes its cue from them. Their reports learn the same posture. The avoidance spreads outward through the structure they sit on top of, and by the time the costs are visible three people have lost their jobs, or a cohort of students has lost a year. The Avoider almost never sees this radiating outward, because from inside their head they are just being principled.

In the book: Chapter 6, where Arjun also names his own version — ten months of deferring to his manager's caution without once forming his own view. "Staying quiet is what juniors do" turns out to be the same trap at a smaller scale.

The asymmetry of the two traps

It is tempting to think of the Dependent and the Avoider as mirror images — one uses too much, the other too little. They are not, and the wrongness of that framing is part of what hides the second trap.

This is why an Avoider with authority is structurally more dangerous than a Dependent with authority. A Dependent boss harms their own career. An Avoider boss harms everyone underneath. Avoidance, when you are alone, is a personal choice. The moment you have anyone reporting to you, it becomes a tax on everyone below.

Type three The Amplifier

The Amplifier uses AI to extend what they already do well. They stay in charge of the thinking. They use the tools where the tools genuinely help, and they don't where they don't. The output gets better and faster — but the person at the centre of it gets sharper, not duller. The skill compounds. They become more, not less.

The four principles

• AI handles the repetitive; you handle the creative. Work that maps cleanly to a process goes to the tool. Work that needs judgement, taste and context stays with you.
• Always verify, never blindly trust. Every output is evaluated against criteria you defined before you saw it. Polish is not correctness.
• Use AI to learn faster, not to skip learning. When you ask the tool to do something, ask it also to teach you. A tutor, not a substitute.
• Your judgement is the product; AI is the tool. What you bring is irreplaceable. What the tool brings is leverage. Never confuse the two.

How to recognise it in yourself

You can defend your work: asked to walk through any line of your most recent piece of work in the next ten minutes, you could — not because you memorised it, because you built it. You evaluate against criteria, not polish, and can spot a plausible-but-wrong AI response within seconds. You iterate without panicking; when something does not work you trace it rather than patch it. And your work is identifiable as yours, because the choices in it — structure, emphasis, what to leave out — reflect judgements nobody else would have made the same way, even where AI produced the prose.

The Amplifier is not a destination. In the book, Arjun reaches it in three months and is placed by the AMQ in the lower-middle band with three named risks. "The Amplifier position is not a destination," Vivek tells him. "It's a posture you maintain across years." The eight-step framework is what maintaining it looks like.

A short self-diagnostic

No scores. Notice your answers, and notice especially the ones where you hesitate; the hesitation is more useful than the answer. For a fuller placement, take the AMQ.

• In the last week, did you press accept on an AI-generated suggestion without reading it carefully? How many times?
• Could you, in the next ten minutes, walk a colleague through the most recent piece of work you produced — every choice, including the parts the tool generated?
• When was the last time you wrote down what good would look like before you saw the AI's output?
• When you read someone else's AI-assisted work, can you tell which parts were theirs? What signal are you using?
• In the last month, did you avoid using an AI tool for a task where it would genuinely have helped? Why?
• When work goes wrong, do you patch the visible thing first, or trace what produced it?
• If you had to teach the next person how you do your job, would the explanation contain any structure — or would it mostly be "I just do it"?

Answers clustering toward fast acceptance, inability to walk through your work, no defined criteria and "I just do it" lean Dependent. Answers clustering toward refusing the tools, defending visibly slower processes on craft grounds and feeling the tools are mostly hype lean Avoider. Most people lean differently in different domains. Notice which.