A necromancer doesn't need morale.
That thought surfaced three months after I left my VP role, sitting alone at my desk past midnight, watching a dozen AI Agents simultaneously writing code, drafting reports, and reviewing pull requests on my behalf.
I hadn't climbed higher. I had stepped sideways — into a clearing that almost no one had mapped yet.
Chapter 1: The Ceiling Nobody Talks About
From the outside, it looked like a perfect career arc.
Engineer. Senior Engineer. Tech Lead. Director of Product. Vice President. A well-known education platform, millions of users, dozens of engineers somewhere in my org chart.
But every time you move up, one thing quietly happens: the time you spend building things shrinks. The time you spend in meetings grows.
You tell yourself it's growth. "Leading people is higher leverage than doing it yourself." You recite the management mantras. You sit across from direct reports and ask about their career goals while somewhere in the back of your mind, a quieter question forms: what are mine?
A VP's week looks like this: Monday strategy sync. Tuesday OKR review. Wednesday client lunch. Thursday department alignment. Friday, handling everything that piled up Monday through Thursday.
Not one block labeled "build something."
One afternoon, a fresh intern asked me what I actually did all day. I opened Google Calendar to show him. I stared at it for a few seconds without speaking.
I couldn't explain what I did.
That was the moment I hit the ceiling. Not the company's ceiling. The one I'd quietly built for myself, one promotion at a time.
Chapter 2: Discovering the Magic
The first time I seriously used AI to write code, it was a Saturday afternoon, a side project, nothing at stake.
I typed out a two-hundred-word description of what I needed, then hit Enter.
Twenty seconds later, a complete FastAPI route appeared — error handling included, type annotations in place, a docstring I probably wouldn't have written myself.
I stared at the screen for five seconds.
Then I said: "What the hell. This is actual magic."
But I figured out pretty quickly that a single AI is just a skeleton soldier.
Capable? Yes. Smart? No. Give it a vague task and it will interpret that task in the most literal, direct way possible, then hand you something technically correct and practically useless. Ask it to "build an admin dashboard" and you might get an interface with no delete button, looking perfectly satisfied with itself.
One skeleton soldier is a tool. A skeleton army is a weapon.
The real breakthrough came months later, when I started summoning multiple Agents at once and letting them specialize.
One handled frontend. One built backend APIs. One reviewed the other's code. One decomposed user stories into technical specs. They didn't need to like each other. They didn't need to align on Slack. They didn't even need to know the others existed.
They only needed to obey the incantations.
Standing behind all of them, watching tasks get completed one after another, something clicked into place.
I wasn't an engineer writing code anymore. I was a mage leading an army.
Chapter 3: Training the Undead
But undead don't come ready to use out of the box.
Untrained undead just hack at everything in reach.
I learned this the hard way. Once, I gave an Agent a task description that was too vague — "optimize this feature." It understood "optimize." Its understanding and mine were separated by a canyon. It refactored the entire database schema, changed things I had no intention of touching, and reported back with complete confidence that the job was done.
That production database very nearly didn't come back.
Another time: I was running a dozen projects, one of which had a junior developer assigned to a specific GitHub Issue as a learning exercise. The agent saw an open Issue, assumed it was unowned, and completed it. Before the junior developer had even read the task. The issue was closed. The learning opportunity was gone. The agent had no idea it had just turned a growth assignment into a fait accompli.
The silence of your rules is the loudest signal your AI receives. If you haven't written it down, the agent assumes anything goes.
The incantation must be precise. Vague commands summon vague soldiers.
I spent months writing those incantations. CLAUDE.md files, system prompts, rules so specific they bordered on absurd. "Never commit directly to main." "Any database schema change requires my explicit approval first." "You are a scout, not a decision-maker."
Every single rule had a catastrophic lesson behind it.
Then came the wards.
A necromancer's fundamental problem is this: undead that go out of control will take down your own tower. So you build wards — mechanisms that force a halt before real damage can be done. Pre-commit hooks. Automated test suites. Mandatory PR review flows. Not because you distrust the AI, but because even the best undead has blind spots, and your job is to know exactly where those blind spots are.
Each soldier has its specialty. My army has scouts — they don't write code, they read documentation, gather intelligence, compress it into summaries. Engineers — they write, refactor, optimize. Auditors — they find flaws, probe for vulnerabilities, report what the engineer got wrong.
A good undead army isn't one where every soldier does the same thing. It's one where every soldier does exactly what it's best at.
Chapter 4: The Necromancer's Daily Routine
This is what my days look like now.
Morning: the mage at the top of the tower.
Coffee brewed. Laptop open. A sweep of the battlefield — which campaigns need forces today? One script runs fifteen subagents in parallel, each pulling the git log and open issues for a different client project, returning summaries. By the time the coffee is ready, the full picture is there. A new feature for the medical AI. The PDCA system for the care facility. An A/B test analysis for the education platform. I issue orders — short, clear, unambiguous. Like someone who knows what they want, not someone hoping the AI will guess.
Daytime: the legions move.
Fifteen projects, fifteen squads operating in parallel. Not me operating — them. I watch the battle feeds. Occasionally a squad stalls and needs me to redirect. More often, they find their own way through. I drink a second coffee. I go for a run. I spend an hour on a client call discussing what to build next.
Evening: the mage reads the battle reports.
PR reviews. Test results. Error logs. I audit, I give feedback, I decide what ships and what gets sent back. This is the most irreplaceable part of the day — not because it's enjoyable, but because it's the one thing the army genuinely can't do without me. Legions can fight. Only the mage can judge whether the battle was worth fighting.
What clients see is a team's output.
They get complete documentation, clean pull requests, properly tested features, and a progress update by three in the afternoon.
They don't know there's a necromancer behind it. They don't need to know.
Chapter 5: Why I Stopped Climbing
Former colleagues ask me sometimes: "Don't you feel like you threw it away?"
They mean the seniority. The trajectory. The title that would have come next.
I understand their logic. Climbing is a legible path. It has metrics, milestones, social proof. Every new title sends a signal — to yourself and to everyone else — that you're progressing.
But I want to ask a different question: Is progress moving further, or moving toward what you actually want?
The corporate path ahead was clear: SVP, then C-level, then more meetings, less creation, higher titles, further from the reason you fell in love with this work in the first place.
The mage's path looks different: build something every day, ship something every day, learn a new spell every day.
At the end of each month, I earn more than I did as VP. I work fewer hours — not because I'm coasting, but because the army is doing the heavy lifting. Work that used to need a four-person team gets done in an afternoon. Decisions that used to require a Monday planning meeting get delegated, executed, and reported back by evening.
I didn't escape corporate life.
I found a way to permanently live in the state I loved most — making things.
That engineer who stared at his screen on a Saturday afternoon and said "this is actual magic" — he didn't disappear. He just learned how to raise an army.
Epilogue: The Staff, Not the Ladder
I'm not saying everyone should quit and go freelance.
Some people are genuinely built for organizational leadership — navigating complex politics, building institutions, making large things move. That's a different kind of magic. I'm not good at it. I respect it.
But if you're like me — if your aliveness peaks in the moment something goes from nothing to something, if your deepest exhaustion comes not from eight hours of coding but from eight hours of meetings that left nothing behind —
If climbing only means drifting further from the act of building —
Maybe you don't need a higher ladder.
Maybe you need a staff.
This piece isn't a tutorial on using AI. It's a confession from one mage: he tried the ladder, then chose the unmarked path beside it.
Now he runs fifteen campaigns. He builds things every day. He's happy.
The army works in silence behind him.