Airmanship: The Habits of a Good Pilot
Module 3 — 75 min + Q&A — outline & speaker notes
Presenter: Jaede · Materials: TBD — Module 3 deck and handout are a future iteration. Audience: anyone who has started using AI regularly — Module 1 (the ladder) assumed, Module 2 (the loop) helpful but not required. Total stage time: 67 minutes of content + 8 minutes of slack = 75, then Q&A.
In aviation, airmanship is the word for everything a license can't test: judgment, discipline, and the habits that keep a competent pilot competent for thirty years. Module 1 showed you the machine. Module 2 gave you the loop. This module is about the habits that decide whether five years of flying makes you sharper — or just more dependent.
0. Cold open (7 min)
- The uncomfortable claim, stated plainly: the tool that makes you faster can also make you worse. Not because it's malicious — because convenience compounds exactly like the loop does. Whatever you stop practicing, you lose.
- The reframe: this isn't a reason to quit flying. Pilots don't refuse instruments; they train so the instruments extend them instead of replacing them. That training is a set of habits, and tonight is the habit list.
- Promise: six habits, each with the failure it prevents, each small enough to start this week.
Speaker notes: This is the most personal module of the three and the one where honesty buys the most trust — you are the person who runs agents at home telling a room that dependence is a real risk. Own it with a first-person story: a task you reached for the machine on and then realized you'd stopped being able to do cold. The room relaxes when the instructor goes first. Don't moralize; the tone is a pilot briefing weather, not a sermon about screens.
Habit 1 — One flight, one destination (9 min)
The habit: one task per session. Start a fresh session for a new task; name the destination in your first message.
Why (the failure it prevents): long meandering threads accumulate confusion — the model's and yours. Its short-term memory fills with abandoned directions, and its answers quietly degrade. Most "the AI got dumb" complaints are this.
Practice this week: before each session, finish the sentence "we're done when ______" — type it as your first line. When the destination changes, land and take off again: new session.
Tradeoff note: fresh sessions cost you re-explaining context — which is exactly what Module 2's logbook and memory files are for. The habits interlock; that's the design.
Speaker notes: Demo-able live in two minutes if you want: show a fifteen-turn wandering thread producing a mushy answer, then the same final question asked clean in a fresh session with one paragraph of context. The clean one wins visibly. Name the principle: the model reads everything in the thread every time — a cluttered thread is a cluttered briefing.
Habit 2 — Trust by stakes, not by fluency (9 min)
The habit: calibrate verification to consequences, not to how confident the answer sounds. Low stakes: accept and move. Money, health, law, reputation: verify against a primary source, every time, no exceptions for answers that "sound right."
Why: fluency is the machine's native talent and your native weakness — polish reads as accuracy to every human ever tested. The failure isn't believing a bad answer once; it's the slow slide into never checking.
Practice this week: write your own two-line stakes policy — "I always verify: __. I never bother verifying: ____." Post it near the screen. The written line is what you'll actually obey at 11pm.
Speaker notes: Reprise Module 1's hallucination card in one breath — this habit is that card turned into a standing policy so the decision isn't remade per-answer. The stakes-policy exercise is the participatory beat of this module; give the room ninety seconds to actually write both lines and ask two volunteers to read theirs. Their examples teach better than yours.
Habit 3 — Fly a leg by hand (11 min)
The habit: deliberately do some tasks unassisted, on a schedule you choose. Draft one email yourself. Read one document cold. Estimate before you ask. Then compare with the machine.
Why: skill you don't use decays — pilots lose hand-flying, and you'll lose drafting, summarizing, and estimating if every rep goes to the machine. The point isn't purity; it's keeping your judgment good enough to check the machine. A supervisor who can't do the task at all can't supervise it (Module 1's checking regimes assume you still can).
Practice this week: pick ONE recurring task you've fully delegated and take back every third rep, by hand. Log how it felt in the logbook — that entry is your skill-decay instrument.
Tradeoff note: this habit costs real time, and for skills you've decided you genuinely don't need to keep, delegating fully is a legitimate call — make it on purpose, in writing, not by drift.
Speaker notes: This is the module's centerpiece and its most contrarian beat — an AI school telling you to use AI less, selectively. Said plainly, it's also the school's credibility: we teach operating the machine, not surrendering to it. The estimate-first variant is the easiest sell for a skeptical room: guess the number before you ask, and you'll feel your calibration either sharpening or rotting. There's no neutral.
Habit 4 — Guard the attention tank (10 min)
The habit: AI time is directed time. You arrive with the destination (Habit 1), you leave when you land. No idle prompting the way you'd idle-scroll a feed; chat has a feed's exact mechanics — variable rewards on tap — and will happily become one.
Why: attention is the fuel every other habit burns. A pilot who fragments their own attention mid-task has defeated every instrument on the panel. (Parents in the room: this is the adult version of exactly what we teach kids about screens.)
Practice this week: notice one moment of recreational prompting — no destination, just pulling the lever — and name it out loud when it happens. Naming it is 80% of ending it.
Speaker notes: Keep this light-handed; the room has heard screen-time lectures and will brace for one. The move that lands is the symmetry confession — "I built an agent so I'd stop checking email, and then caught myself checking the agent" — laughter, then the point. One sentence of the anti-brainrot thread from the tutoring side of the house is enough; adults recognize it applies to them the moment it's said without judgment.
Habit 5 — Know what it costs (8 min)
The habit: a monthly five-minute instrument scan of the money: what subscriptions exist, what usage-billing exists, what each one actually earned you this month. Cancel one thing a quarter if the scan can't defend it.
Why: AI spend is designed to be ambient — small monthly numbers that never re-justify themselves. The failure isn't one $20 subscription; it's the drawer full of them, plus the usage bill nobody watches. Rung 5 builders: usage billing without a monthly look is how hobby projects grow silent triple-digit months.
Practice this week: list every AI-adjacent charge from your last card statement, next to what it did for you. Five minutes. Keep the list in the logbook.
Speaker notes: Shortest habit, deliberately placed after the heavy ones — the room needs a concrete, unemotional win here. Real numbers work: name what your own stack costs monthly and what you cancelled last quarter and why. The "cancel one thing a quarter" rule gets written down more than anything else in this module; make sure it lands verbatim.
Habit 6 — Keep the logbook honest (6 min)
The habit: log failures with the same care as wins — what you asked, what went wrong, what you'd brief differently. One honest line beats five victory lines.
Why: Module 2 taught the logbook as memory for the system. Airmanship uses it as an instrument on you: your drift, your dependence, your calibration only show up in a record honest enough to show them. Aviation figured this out generations ago — the safety culture is built on pilots filing reports on their own mistakes.
Practice this week: add one line to your logbook template: "What I'd brief differently." Fill it in whenever a session lands somewhere you didn't intend.
Speaker notes: Close the loop explicitly across all three modules: the ladder told you what the machine can do, the loop makes every session smarter, and airmanship keeps you worth compounding. End on the aviation line — every improvement in flight safety was written by somebody's honest logbook — and stop there. It's the right final image for the course.
Close (7 min)
- The six habits restated in one pass, one line each (slide).
- The homework, restated as a single sentence: pick two habits — not six — and run them for two weeks; the logbook will tell you what they changed.
- Where the materials live, where the list lives: handsinjars.com/pilot. No dates promised.
Speaker notes: Two habits, not six, is load-bearing — six new habits is zero new habits. Let the room pick out loud if it's small enough; hearing neighbors choose differently makes the spectrum point (Module 0's "self-defined milestone") without re-teaching it. Q&A follows.
Q&A posture (notes only)
- "Isn't this all just self-discipline stuff?" — Yes, and so is most of flying. The machine's capabilities are the exciting part; the habits are the part that determines whether you still have your own capabilities in five years. Every tool this powerful has an operating discipline. This is ours.
- "Which habits matter most?" — For heavy users: one-flight-one-destination and fly-a-leg -by-hand. For light users: trust-by-stakes alone repays the evening. When in doubt, the logbook habit — it's the one that improves the others.
- "Is this the last module?" — It's the last one that exists tonight. The course grows from what students actually hit — that's the loop, applied to the school itself.
Open source under CC BY-SA 4.0. · AI Pilot School