The Practice Loop

"A thousand reps with a blurry mirror just makes you very good at your own mistakes."

A Thousand Portraits

Loomis exists because of a promise Dad made to himself: a thousand portraits, then a thousand poses, drawn on purpose. Not a thousand doodles — a thousand deliberate constructions, each one an attempt to see a face or a figure as a solid in space and get the angles right.

Here is the problem volume alone can't solve. When you draw alone, who tells you the jaw was five degrees too steep? You can feel that something is off without knowing what. Repeat that for a thousand drawings and you don't get a thousand times better — you get very consistent at the same small errors, because nothing ever corrected them.

The Loop, in Four Beats

Deliberate practice is a loop, and it only works when every beat is present:

Inspect. Study the reference as a 3-D structure, not an outline.
Draw. Construct it yourself — the ball, the box, the axes — before you look at any answer.
Reveal. Loomis shows the deterministic construction it recovered from that exact photo.
Compare. Lay yours beside the answer and find the drift. That is where the learning actually happens.

Loomis owns exactly one of those beats — reveal. The other three are yours. The engine is not the practice; it is the honest mirror the practice was missing.

Why the Reveal Comes Second, Never First

Order is not a detail here. If the answer appears before you draw, your hand quietly copies it and you learn nothing — you traced. The whole value is in committing to your own read first, being wrong on the record, and only then seeing by how much. That is why Loomis refuses to draw for you, and why the grader (Track 6) treats your attempt as append-only: your guess is preserved exactly as you made it, before any answer touched it.

Feedback Quality Is the Multiplier

Reps matter, but they multiply against feedback quality. Slow, vague feedback caps how fast you improve no matter how many hours you put in. A sharp, immediate, reproducible answer key turns each rep into a real correction. That multiplier — not the drawing tool, not the reference library — is what an art-study engine is really for.

Code

The deliberate-practice loop, as pseudocode·text

for reference in deck:              # a thousand of them, on purpose
    study(reference)                # YOU: read the structure, not the outline
    my_construction = draw(you)      # YOU: ball, box, axes -- commit before looking

    answer = loomis.reveal(reference)  # ENGINE: the deterministic construction

    diff = compare(my_construction, answer)   # YOU: where did the angle drift?
    remember(diff)                   # tomorrow's rep starts from a corrected eye

# The engine owns ONE line. The other four are the practice.
# Delete the 'draw before reveal' order and the whole loop becomes tracing.

Exercise

Pick a skill you're actively grinding reps on — drawing, a language, an instrument, coding. Write down your current feedback signal: after each rep, what tells you whether you were right, and how fast? Then rate it: is it honest, immediate, and reproducible — or slow, vague, and self-graded? Design one change that would sharpen it.

Hint

If your only feedback is 'it felt okay' or 'I'll notice eventually,' you're practicing with a blurry mirror. The fix is usually an external, reproducible reference you compare against right after each attempt — which is exactly what Loomis is for drawing.