Contents
The trainer has six typing modes. Harder is not always better — each mode removes a different crutch. The Train button on a blog code block always opens full; you can switch modes in the setup panel afterward.
What an exercise is
- Pick language, level, and mode (plus sliders when needed: how much to hide, how long to show the reference).
- Load a bank example, generate one, or arrive from a blog block with the code already filled in.
- Type in the input. Characters are checked against the reference: green = match, red = mistake.
- The drill ends only on a full match. Then you see stats, the next example, or a link back to the article.
Panel metrics: accuracy, WPM, mistakes, time. For learning, a clean finish beats a WPM spike with typos.
full — first pass
The whole reference stays visible. You type alongside it.
When: first time with a snippet; unfamiliar syntax; longer blocks where structure matters.
How: one calm pass to 100% accuracy. If you stall on a name, reread that line aloud, then continue.
partial — holes in the reference
Some lines are hidden (slider sets the share). You reconstruct them from neighbors.
When: you already did full; you want to check import → function → return links.
How: try 30–50% hidden. Guessing often? Back to full. Too easy? Raise the hide ratio or move to memory.
memory — from memory
The reference shows for N seconds, then disappears. You type what you retained.
When: short snippets; preparing to write similar code without the article open.
How: start at 15–20 seconds. Do not peek at the article tab. End-of-run mistakes show what did not stick.
fillblanks — fill the gaps
Key pieces are cut out of the reference. Restore them exactly.
When: you know the shape but mix up API details.
How: one pass; note which gaps hurt — a one-day note, not a forever wiki.
hints — skeleton
You see a scaffold and hints, not a full reprint list. Closer to “rebuild from a contour.”
When: full/partial already feel solid; bridging toward your editor.
How: finish a try before opening the real file in a repo. Then diff.
implement — from a prompt
You see the task text, not the solution. Matching is against a stored reference (strict character match).
When: bank exercises with a description; checking whether you can reproduce a pattern cold.
How: do full on the same snippet first, then implement a day later. Exact match is harsh — use it on short canonical snippets; otherwise return to partial/memory.
A weekly ritual
| Day | Mode | Goal |
|---|---|---|
| 1 | full | 1–2 article snippets to 100% accuracy |
| 2 | partial | same snippets, ~40% hidden |
| 3 | memory | short blocks, 10–20s preview |
| 4 | fillblanks or hints | API details / scaffold |
| 5 | full from a new post | grow the set |
Use Train on a blog block to jump in with code ready. Bank + mode switching live on the trainer page.
WPM, accuracy, leaderboard
- Accuracy is the learning signal. Aim for 100% per pass.
- WPM is secondary; it rises when brackets stop feeling foreign.
- Mistakes matter by where (indent, quotes, names), not only count.
- The leaderboard blends speed and accuracy; set a nick in the progress widget. Motivation, not an exam.
Also: landing page, reading code posts, leaderboard.
Comments