The AI-Native Engineer

Reference

Glossary

The words the frameworks use, each in a sentence or two, with a link to the framework where it does its work.

AI-assisted
Using the tools to do your existing work faster. It is real and worth having, and it is where most working engineers currently are.
From AI-Assisted vs AI-Native
AI-native engineer
An AI-native engineer is not the person who uses AI most. It is the person who compounds judgment through AI: knowing, before the handoff, which parts of the work carry judgment that lives only in you, and refusing to price those parts at the commodity rate.
From AI-Assisted vs AI-Native
The announcement layer
Everything you hear about how your field is changing: launch posts, conference demos, a colleague's promotion. All of it is real information about outcomes and almost none of it is information about method.
From The Honest Look
Approved vs reviewed
Approved and reviewed are not the same word. Four lines you understood and twenty-six that looked fine is a normal week now, and the distance between those two words is where the risk of this decade lives.
From The Three Kinds of Work
Asking for the decision
Judgment cannot be assessed by asking someone about a system. It can be assessed by asking for one hard call they personally made on something that mattered, and then asking again.
From Asking for the Decision
Catching
The machine produces something plausible and someone has to notice what is wrong with it. This is new work. It is thin per instance and it accumulates, and almost nobody budgets for it.
From The Three Kinds of Work
The cheap years
The queue of low-stakes work that trained the people above you. It is being automated, sometimes by you personally, and the only substitute anyone has found is contact.
From The Three Seats
Contact knowledge
Knowledge that genuinely cannot be written down, as opposed to stated knowledge. If it is contact knowledge, name who else is getting that contact. If the answer is nobody, that is a staffing risk with a date on it.
From What Cannot Be Delegated
The decision entry
Four lines written before you answer a real decision: what you are choosing, what it costs, what would prove you wrong, and when you will check. Written after the fact, this is a justification. Written before, it is the difference between a choice and evidence.
From The Three Traps and the Decision Entry
The delegation ladder
Delegating work to an AI system is not one decision about whether to trust it. It is a ladder of five rungs, from the machine typing while you think to the machine owning the outcome and reporting only exceptions.
From The Delegation Ladder
The handoff line
The line between work that has become safe to hand off and work that has not. It moves without announcing itself. The failure is almost never the act of handing off; it is not noticing when a handoff has stopped being safe.
From The Handoff Line
The honest look
Finding out how the work actually gets done now by someone doing it effectively, close enough to see the mess. The gap between you and the person ahead of you is usually a working loop, not a talent.
From The Honest Look
The passenger
Hands the whole job over and keeps the title. Nothing is accumulating, and there is no judgment left in the loop to notice when the output stops being right.
From The Three Traps and the Decision Entry
The portfolio of decisions
Judgment is invisible until it is made legible, and the evidence is decisions rather than output: what you chose, what it cost, what you turned down, and how you found out you were wrong.
From The Portfolio of Decisions
The purist
Refuses the tools on principle. The judgment stays intact and the scope keeps shrinking, and the argument for the position gets more articulate every year as the ground it stands on gets smaller.
From The Three Traps and the Decision Entry
The record
A written record of your own reasoning, sitting where the work happens. One rule you own, with the reason written above it, your name and the date on it. Then again next week, until the record is worth more than the output.
From AI-Assisted vs AI-Native
Rung
One of the five levels of the delegation ladder: typing, drafting, task, direction, outcome. No rung is the correct one. The work sets the rung, and two questions decide which: how much of the judgment is written down where the machine can read it, and how expensive wrong is and how fast you would find out.
From The Delegation Ladder
The seam
Where the work happens: the code, the fixture, the ticket. A note in a document nobody opens does not move the line. A note at the seam does, because it is where the next receiver, human or otherwise, is already looking.
From The Handoff Line
The three seats
The shift does not price every career stage the same way. Early, the cheap years that used to train you are being automated. In the middle, your slow years are worth more than your output. Senior, most of your value is multiplying other people's judgment, which cannot be verified from the outside.
From The Three Seats
The tourist
Adopts everything the week it ships and stays with nothing long enough for a record to form. The hardest one to see from inside, because it feels like being current.
From The Three Traps and the Decision Entry
What cannot be delegated
Not all valuable knowledge becomes more abstract as tools improve. Some of it becomes more specific, tied to one system's history and to people who have watched it fail.
From What Cannot Be Delegated
Work that got cheap
Boilerplate. Test setup. A validation layer. Schema plumbing. Real work, formerly yours, now available at a much lower price.
From The Three Kinds of Work
Work that is still yours
Knowing where to start. Knowing when the question itself is wrong. Knowing that one real case in production outweighs a clean theory about all of them.
From The Three Kinds of Work