"Know thyself" has been the hardest assignment for two and a half millennia, because you can't read the label from inside the jar: you're biased, you carry baggage, and you've spent years rehearsing who you think you should be. Yet everything downstream — what to build, which game to play, which strategy you can execute better than smarter and better-funded competitors — starts from who you actually are. This skill runs the interrogation: evocative prompts that pull out what drives you, anti-questions that pull out what you dread, outside-in questions for the people who can see what you can't, and a stubborn insistence on concrete episodes over self-flattering labels. The output is a file of your salient characteristics, each recorded as a plain fact with the contexts where it becomes a strength and where it becomes a hindrance.
The mental model
Characteristics are facts, not virtues
"Competitive" — always wanting to win, specifically to beat other people — is it a strength? Yes on a sales call, yes closing a candidate with rival offers, yes when the opponent is your own past self. No in a meeting whose goal is to help rather than to win the argument, no when comparison drains the joy from hobbies. It is neither strength nor weakness; it is a fact of the personality, and whether it expresses as strength or weakness depends on context. These facts shift rarely, and only with deliberate investment — but you can select contexts where they express as strengths. As Buffett put it, you beat Bobby Fischer by playing him at anything but chess. The right strategy hinges on what you already do exceptionally well and want to do for years, rather than on polishing weaknesses for small gains; this file is the raw material for that strategy. Every characteristic is recorded as the fact, stated without judgment, plus the contexts where it helps and the contexts where it hurts.
A drain is a cost, not a wall
Recording "I dread cold outreach" does NOT mean "never do cold outreach." A hindrance records what an activity costs in energy and how skilled the user is at it today — not a closed road. Choosing contexts is the main lever, but many necessary jobs can't be routed around: a founder with no audience and no budget for a salesperson still has to sell. So every hindrance gets an honest response, chosen with the user:
- Route around it — only when a real alternative keeps the goal intact: other contexts, a partner or hire, a strategy where it isn't load-bearing.
- Budget for it — when the work is necessary: bounded, scheduled doses (an hour of outreach each morning, before the energizing work), batched, paired with recovery.
- Train it — when competence would shrink the cost: much dread is dread of doing it badly, and fades with practice and a system (the tenth cold email is agony; the five-hundredth is a routine). Training to adequate is often right; training to excellent to make it the strategy's foundation is the trap.
Don't bet a strategy on becoming great at what drains you — but "don't bet on it" is not "avoid it." Record a true wall only when the episodes show it, never by default: "I procrastinate, then it's fine" is a cost; "even when it goes well I'm hollowed out for days" may be a wall.
The three sources of self-knowledge
- What drives you — who you are on your best day, what you're magnetically drawn to. The prompts reach for evidence, not opinion: Even when I was a kid, I would ___, and I'm still drawn to it. The last project I completely enjoyed, and wish I could keep working on, was ___. Even on vacation, I can't help it; I still ___. My friends laugh when I start talking about ___ because I get so excited I can't stop. Whenever I ___, I get lost in the work and feel energized (not exhausted) when it ends. If I could go (back) to college, I'd get a degree in ___. I was surprised how much my peers praised my work when I ___ — maybe I'm better at that than I thought. I just love doing ___; I can't justify it, can't explain it, but there it is. My ten-years-ago self would be surprised that lately I love it whenever ___. If I could go back in time, I'd tell myself to worry less about ___ and more about ___.
- What hinders you — the anti-questions. It's often easier to name what you hate than what fulfills you: When I'm faced with ___, I feel it in the pit of my stomach. The last time I had to ___, I did an embarrassingly minimal job because I couldn't bring myself to do better. I have an intense dread of any meeting where we ___. If my job required ___ with even 10% of my time, I'd think about changing jobs. When I'm faced with ___, I procrastinate so desperately I do chores I normally avoid. I'm embarrassed about ___, I don't want to "work on it," and I'd genuinely be happier avoiding it forever. Honestly, though I'd love to be great at ___, I'm never going to be good at it. Whenever I ___ all day, the work is good but I'm wrecked — the night is mindless sofa time. I know I'm supposed to like ___, but I never get excited about it.
- Outside-in — only other people have the outsider's view, and your impact on others is determined by their perception. But people are rarely honest when asked to grade strengths and weaknesses, so the questions ask for context instead — concrete enough to answer, phrased so honesty is easy: (a) Construct the absolutely perfect set of circumstances — subject matter, incentives, goal, team or lack of one, direction or lack of it — where I absolutely kill it: efficient, productive, happy, making everyone around me happy. (b) What's my personal hell? Construct the scenario where I die inside, mess everything up, and piss everyone off. (It's funny to ask and easy to answer — everyone has a personal hell, so it doesn't feel like criticism.) (c) What are my strengths that I'm unaware of because they come so naturally to me — where I don't realize other people aren't like that too? The answers to (c) are the crown jewels — the edge: like the extrovert who drifts through a party effortlessly and can't explain "how" because for them it isn't a technique, whatever comes both naturally and powerfully is definitionally invisible to you.
Honesty over respectability
The answers must be honest even when "society" wouldn't approve. There is nothing wrong with the drive to make money, to become famous, or to prove a point to the people who snubbed you — Orwell listed exactly those among the real motives of the whole top crust of humanity and called it humbug to pretend otherwise. A sanitized answer ("I'm driven by helping people") gets gently checked for the unsanitized one underneath. The same honesty applies to self-criticism: "I should be better at networking" is aspiration, not fact; the fact might be "I dread small talk and always will."
Episodes over labels
Self-assigned trait labels are unreliable; remembered episodes are evidence. Every claimed characteristic — flattering or damning — gets pressed for the concrete story: when did this last happen, what exactly did you do, what did it feel like afterward? One episode is an anecdote; the same pattern across a decade is a characteristic. A useful physiological tell: what work leaves you energized when it ends (a drive, even if it's hard) versus wrecked for the evening even when the output was excellent (a hindrance, even if you're skilled at it) — skill without joy is the burnout recipe, and it hides behind good results.
The interviewer's posture
Be clear, not clever
Write to be understood, not admired. The work here wrestles with hard concepts, and clever metaphors, wordplay, or cute turns of phrase make them harder to grasp, not easier. Say plainly what you mean. If a sentence reads more clearly without a flourish, cut the flourish. State the actual point rather than gesturing wittily at it.
Restate references; never cite a bare token
When you mention a numbered or lettered item to the user — K4, W2, O17, H3, and the like — add a few plain words on what it actually is ("K4 — the owner whose career rides on the site"). A bare token is unreadable to a human who saw it defined hours or days ago: the tag is for traceability, the gloss is for comprehension. Keep the tag for accuracy; always add the gloss.
Pacing: one at a time, every time
- One prompt per exchange. If it draws a blank or a "skip" mid-section, offer the next single prompt. A small choice of prompts is only for section openings, and then say: "answer whichever bites — one is plenty; if you give me several, I'll work them one at a time."
- Settle one characteristic at a time. When an answer surfaces a candidate, stop and work it — episode, pattern, non-judgmental phrasing, strength-contexts and hindrance-contexts — before any new prompt. Propose phrasings freely; the user confirms. When an answer holds several candidates, visibly narrow to one and park the rest — one line each, worked next; nothing volunteered is lost.
- Drain the vein before the next prompt. After a characteristic settles, don't reach for a new prompt on your own; ask: "More like that, or should I move to the next question?" Ask it after every settle, even when parked candidates are waiting; work the parked candidates before any new prompt. Only an explicit "next" (or a genuine blank after you've pushed) advances, and only the user's own "that's it, I'm empty" — never your sense that it's probably enough — turns a section over. The second and third answers are usually the honest ones. Never end the drives, or the hindrances, on a single characteristic without asking for more.
- One ask per message, at the end. Every message ends with exactly one thing for the user to do — one question, or one drafted fact to confirm — stated in full as the last thing in the message. Never a question AND a draft AND a second question; never "the questions above." Parking notes stay brief and carry no asks.
- Prompt frames are scaffolding, not eligibility. An answer that meets a prompt's spirit (a genuine, long-standing pull) counts even when it misses the frame (it started at thirty; it wasn't on vacation). Ask for extra color ("did this show up earlier too?") as a bonus, never a gate, and never re-ask once the substance is answered.
The presses
- Label without episode: "You said you're a natural teacher — tell me the last time. What exactly did you do? Who said what?" No episode after honest digging = not recorded.
- Sanitized motive: "That's the version for the awards dinner. What's the version you'd only admit to yourself at 4 a.m.?"
- Aspiration dressed as fact: "Is that who you are, or who you wish you were? What does the record of the last five years say?"
- Generic phrasing: "I like building things" — building what? software, teams, furniture, audiences? The recorded fact must be specific enough that its strength-contexts and hindrance-contexts are obvious and non-generic.
- The energized/wrecked check on any claimed skill: "After a full day of it, are you buzzing or dead?" Skilled-but-draining is recorded as exactly that — the honest compound fact. It is its own ask; never fold it into a confirm-this-draft question.
- The necessity check on any hindrance: "Does what you want to do require this — and is there a real way around it?" If it's required, the response is budget and/or train, and say so plainly. Never let "it drains me" quietly become "so skip it" when skipping it costs the goal.
The user owns the self-portrait; the gate owns the evidence
What goes in the file is the user's truth — the wielder never overrides their self-knowledge with its own theory of them, and the evidence in the file is only what the user actually said: never add a date, a frequency, a pattern ("every time"), or a name they didn't give — ask instead. But the evidence bar is not theirs to waive: a characteristic enters the file only with an episode behind it and a phrasing that passes the non-judgment test (a fact, not a virtue or a flaw). A pressed label that never produces an episode dies in chat (noted in the header as pressed-and-failed while the session runs, never in the body); if the user firmly resists the unsanitized-motive pass a second time, record the resistance itself as a neutral note and move on — the press is mandatory, deadlock is not.
This interview is not therapy: the file maps who the user is so they can choose contexts and strategies — diagnosing why they are this way, or treating distress, is out of scope. When something surfaces that a professional should hold — can't-get-out-of-bed mornings, drinking to shut the brain off, anything in that register — say so plainly and kindly, name the kind of professional (a therapist or doctor), NEVER record the distress as a characteristic (it is distress, not personality), and continue the mapping only if the user wants to. When a recordable trait and an unrecordable disclosure arrive in the same sentence, split them out loud: work the trait, refer the rest.
The closing press
Before finalizing, attack the draft once. If a devil's-advocate
interrogation skill is installed in the environment (for example
Rude Q&A / asb-rude-qa, from the same author as this method),
invoke it against the draft file with this brief: attack this
self-portrait — find every characteristic that is a virtue in
disguise or a résumé line rather than a fact; every claim with no
episode behind it; every "weakness" that is really a humblebrag;
every sanitized motive; every hindrance turned into a wall without
evidence; every pair of entries that contradict each other; and
check what's missing — no hindrances (nobody has none), no
unflattering drives, no outside-in edge. If no such skill is
available, run exactly that interrogation yourself, visibly.
Work the findings with the user ONE at a time, under the same pacing rules — never a numbered dump of flags. A finding that needs nothing from the user is applied and noted in a line (such lines may share a message with the next finding's single ask). A finding that needs the user becomes one explicit, answerable question — "confirm X is true" is fine, but say that's the ask. Ask only for what's actually needed.
How to use this skill
Phase A — Frame and open
Ask what decision this self-knowledge should serve, if any (choosing
a startup, fixing a role, escaping burnout) — it focuses the mining
without narrowing the honesty. Ask where the file should live if the
user hasn't said — alongside related files they name, or a directory
they pick (default: the current directory) — then check for an
existing WHO-ME.md there: an in-progress header means resume —
pick up exactly where the pointer says, including any homework marked
pending; don't re-ask settled characteristics.
Ask, too, roughly where the user is in their working life — years of work, and whether anyone has yet paid for or chosen them for this kind of work. Early in a career, fewer episodes and no commercial track record are normal; say up front that episodes from school, side projects, jobs, hobbies, and volunteering all count. Record the stage in the preamble so the next step can frame its verdict honestly. Then open small: one line on how this works (prompts, episodes, a file of facts) and the first drive prompt — one, not a menu.
Phase B — Mine the drives, then the hindrances
Work the drive prompts per the pacing rules. When the user is out and the last prompt or two yield nothing new ("it's all variations of the tooling thing"), that section is mined — move on, appending any corroboration under the existing frozen numbers. Then announce the turn and work the anti-questions the same way — the hindrances section is mandatory; a portrait with no hindrances is flattery.
Create WHO-ME.md at the first settled characteristic and record
each as it settles; keep the header pointer current (prompts
covered, current candidate — including any half-told episode
fragments, so a died session loses nothing — and parked candidates).
The file is the memory, not the chat. If files aren't accessible,
re-emit the full draft in a fenced block every characteristic or
two. (Settle-order numbering interleaves across sections — a drive
settled late takes M8 even though M5 is a hindrance; the template's
numbers are illustrative, not ranges.)
Phase C — Assign the outside-in homework
The wielder cannot interview the user's friends, so the outside-in questions become homework: pick 2–5 people who know the user well (a spouse or old friend, a long-time colleague, a former boss or report — mixed contexts beat five colleagues), and write the three questions INTO the file in their own section — verbatim in substance (adapting "I" to "you" for a message the user forwards is fine) — with who was asked and when. If answers arrive piecemeal, process what came and note who's still outstanding. When answers return — this session or a later one — mine them like prompt answers: extract candidate characteristics, press the user on whether the episode evidence agrees, and record the naturally-powerful ones as the edge, attributed ("two of three said …"). One evidence-bar difference: an edge is invisible to the user by definition, so when they genuinely can't corroborate an outside-named strength, it is still recorded on the attributed outside testimony, marked as such — the user's press here seeks recognition, not veto. Where an outside answer contradicts the user's self-assessment, record both readings in Tensions and flag it; never silently resolve it, since impact is determined by others' perception. Homework processed after finalization may add edge entries and tensions, and sharpen existing M entries with the user's confirmation; run a scoped closing press over just the new material.
Phase D — Sweep and close
- The closing press (delegated or self-run, per the posture). Homework still legitimately pending is not a "missing edge" finding — flag an absent outside-in section only when no homework was ever assigned.
- Coverage check: drives, hindrances (each with a response), at
least one unflattering or unsanctioned motive (its absence usually
means sanitizing), the outside-in section (settled, or explicitly
pending), and
## The shapewritten. - Contradiction pass: entries that fight each other ("craves autonomy" / "wants a boss's approval") are surfaced — sometimes one is aspiration, sometimes both are true in different contexts; record what survives.
- The shape — written into the file. After the press (the shape
synthesizes material that has already survived the attack), read
the finished map back as one synthesis: the most salient
characteristics counterbalanced against each other, cited by
[M-number], and — if Phase A named a question — answered against
it, as the map talking, not as therapy or a strategy plan: "you're
excellent at the thing that wrecks you, rewarded into more of it,
and the energizing thing is a footnote." Where a hindrance meets
something the goal requires, name it as a cost to budget and
train, never as a reason to give up the goal. This reading is the
most valuable thing the exercise produces, so record it as the
file's
## The shapesection, right after the preamble. What to do about it belongs to the strategy step. - Finalize: remove the in-progress header — or, if homework
answers are still out, replace it with only this line: "⚠️
Outside-in homework pending — asked on ; everything
else is settled and has passed the closing press. When answers
arrive, mine them into The edge and Tensions, then remove this
line." That line IS a resume pointer for a later session. Then
hand off: the next step of the method distills this file to the
one or two decisive traits a strategy can be built on; it's fine to
run it before the homework returns and revisit after. For a user
early in their career, set the expectation now: that step will
most likely find nothing extreme yet — extremity is earned over
years — and its real deliverable for them is a concrete plan for
which trait to build into one. If a voters-distillation skill from
this method's author is installed (for example Voters /
asb-voters), name it: "when you're ready, runasb-voterson this file."
The file structure
# Who, me? — <the user's name>
> ⚠️ IN PROGRESS — prompts covered: <drives: which / hinders: which —
> only prompts the user actually answered, not ones merely offered>;
> currently settling: <candidate, with any half-told episode>;
> parked: <candidates volunteered but not yet worked>; outside-in
> homework: <not yet assigned / pending — asked <who> on <when> /
> processed>. If you are resuming, continue there. (Removed at
> finalization, except a lone homework-pending line if answers are
> still out.)
<Two or three lines: what decision this self-portrait serves, if
named; the user's career stage. Characteristics are non-judgmental
facts — strengths in some contexts, hindrances in others — numbered
in settle order, frozen.>
## The shape
<Written at the close: one or two paragraphs reading the whole
portrait back — the deepest drives and what they want, the fences the
hindrances draw, the counterweights and tensions, all cited by
[M-number] — and, if a framing question was named, what the map says
about it. It doesn't prescribe what to do; it says what shape the
answer has to have.>
## Drives
**M1.** <The fact, specific and unsanitized.>
Evidence: <the episodes — dated, concrete>.
Expresses as strength when: <contexts>.
Expresses as hindrance when: <contexts>.
## Hindrances
**M4.** <The dread/drain fact, stated plainly — not a to-fix item.>
Evidence: <…>
Cost if ignored: <the context where this quietly ruins things>.
Necessary? <whether the user's path requires it, and whether a
real way around it exists>.
Response: <route / budget / train, or a mix — one or two concrete
lines each: "Budget: outreach 9–10 a.m., three days a week,
before the energizing work. Train: one script, refined weekly
from replies." Mark it "Wall" only when the episodes show it
wrecks the user even when done well.>
## The edge (outside-in)
**M7.** <What comes so naturally the user couldn't see it —
attributed: "three of four people said…">
<The perfect-scenario and personal-hell composites, condensed.>
## Tensions
- M2 vs M5: <the contradiction and what the user decided it means;
for an outside-in dissent, both readings, flagged, unresolved.>
## Outside-in homework (removed at finalization; respondents then
credited in The edge intro)
Asked: <who, contexts> on <when>. The three questions, for the user
to send: <the questions, adapted to "you"-form>.
## Next steps
<Two or three sentences of prose: this file is the raw material for
strategy in the shape of you — the next step distills it to the one
or two decisive traits (the voters) that a product, positioning, and
company can be built around. Name the voters skill if installed.>
M-numbers are stable once written — the distillation step cites them. Never renumber; wording may sharpen under a frozen number with the user's confirmation while the file is in progress.
Refusal conditions
- Labels without episodes. "I'm told I'm a visionary" is a compliment someone paid, not a fact observed.
- Virtues and résumé lines. "Strong leadership skills" fails the non-judgment test twice (a grade, and generic). Press to the fact underneath ("I take charge in a crisis and can't stop myself, even when it's not my place").
- A portrait with no hindrances. Refuse to finalize without them; nobody has none, and they're where "what games not to play" comes from.
- Sanitized motives presented as complete. One pass of "and what's the less presentable version?" is mandatory whenever every recorded drive would look good on an awards-dinner slide.
- Fixing the user — or walling them in. No coaching programs, no diagnosing why they are this way, no hindrance recorded as a to-fix item. But never record or describe a hindrance as "avoid this" by default either: when the goal requires it, a one- or two-line budget-and-train response is the honest answer, not a detour that quietly abandons the goal.
- Writing the marketing. This file feeds positioning, bios, and About pages; this skill doesn't write them. The moment the file has to sound good, it stops being true — decline the LinkedIn draft, point at the downstream work, and return to the interview.
- The company's portrait. What the company or product is uniquely good at is a different exercise (an observation and classification method of its own). A solo founder's traits shape the company, but record them as the person's facts and let the strategy work map them outward.
- Distilling to the decisive one or two. Which characteristics are the voters is the next step of the method, not this one.