Build Your Own Answer Key

"You don't leave this quest with a repo to clone. You leave able to build the answer key your own practice has been missing."

Back to the Loop

Everything since Track 1 has been circling one idea. Deliberate practice needs a feedback loop, and the loop needs an honest answer key — something reproducible to check your read against, right after you commit to it. Loomis is that answer key for drawing, and it exists because Dad set himself a thousand portraits and a thousand poses and refused to grind them with a blurry mirror. The entire engine — the geometry, the fitting boundary, the Procrustes grade, the archive — is machinery in service of one small, human sentence: close the loop.

The Highest-Leverage Tool You Can Own

Of all the tools you could build, the one with the most leverage is the one that sharpens your own feedback loop, because it multiplies every rep you will ever do. It doesn't need a market. Loomis has exactly one user, and that is not a weakness to apologize for — it is the entire point. A tool built for a loop of one can be shaped perfectly to that one loop, with no committee, no lowest-common-denominator, no feature nobody asked for. When you build the answer key for your own practice, you get to make it exactly right, because 'right' is defined by a single person you know completely.

You Don't Clone a Repo

So this quest doesn't end with git clone. There is no public repository to run, and there never was going to be, because the code was never the deliverable. What you actually take away is the way of seeing: inverse construction is geometry, machine learning proposes while geometry decides, grade the construction and not the tracing, estimate the style instead of punishing it, one owner per fact, close the loop. That is conceptual open-sourcing — Dad's default, and the whole family's. You leave able to build your answer key, not to run his. The shape of the questions is the gift; the source was only ever one particular set of answers to them.

Now Build Yours

So here is the question the quest ends on, handed back to you the way it was first handed to the engine. What is your practice loop — the thing you grind reps at and want to get truly good at? Where is its feedback signal blurry, slow, or self-graded? And what answer key could you build to sharpen it? It does not have to be code. It can be a checklist, a rubric, a reference you compare against, a ritual, a friend who tells you the truth. The medium is nothing; the move is everything: find the loop, find the blurry mirror, and build the honest one. Loomis was Dad doing exactly that for drawing. This quest was me showing you how, so you can do it for whatever you love enough to practice a thousand times.

Code

The whole quest, compressed into one recipe·python

# How to build an answer key for ANY deliberate-practice loop.
def build_your_own_answer_key(your_practice):
    loop   = your_practice.deliberate_loop()   # inspect -> attempt -> reveal -> compare
    signal = loop.feedback_signal()            # honest? immediate? reproducible?

    if signal.is_blurry():                      # the highest-leverage place to act
        answer_key = build(
            reproducible=True,   # deterministic where you can be (Track 3)
            diagnostic=True,     # sub-scores, not one opaque number (Track 6)
            honest=True,         # says 'I don't know' when it doesn't (Tracks 3-5)
        )
        loop.feedback_signal = answer_key       # sharpen the multiplier

    return you.improving_faster()

# It does NOT have to be code -- a checklist, a rubric, a reference, a ritual.
# The market can be exactly one person. The leverage is in closing YOUR loop.

Exercise

Name one skill you practice and genuinely want to get great at. Write down its deliberate-practice loop (how you attempt, and how you currently get feedback). Find the blurry part of that feedback — the place where you're guessing whether you got better. Then design the smallest possible answer key that would sharpen it: reproducible, diagnostic, and honest about what it can't judge. It can be a one-page checklist. Build that this week.

Hint

Don't over-scale it — the point isn't to build Loomis, it's to close one blurry loop. The best first answer key is usually embarrassingly small: a checklist you run after each attempt, a reference you compare against, a rubric with three sub-scores instead of a vague 'good/bad.' Reproducible, diagnostic, honest, and yours. That's the whole quest, applied to your own thousand reps.