The day desktop typing quietly changed
For 41 years, Notepad was the place you went when you wanted nothing between you and plain text. That changed in early July 2024, when Microsoft’s rollout made spell check and autocorrect standard in Notepad for Windows 11—a milestone that effectively turns even the bare‑bones editor into an as‑you‑type fixer. (arstechnica.com)
Why does that matter to typing tests and training? Because the moment your editor corrects you, old assumptions about “raw” speed and accuracy start to wobble. If your typos are silently patched, your WPM goes up and your visible error rate goes down—even if your underlying precision didn’t change.
What exactly changed in Notepad (and how to toggle it)
Microsoft’s own announcement lays out the details: Notepad now underlines misspellings, offers suggestions, and includes autocorrect that “seamlessly fixes common typing mistakes as you type.” It supports multiple languages, and you can enable or disable these features globally, per file type, or just for the current document. Notably, the features are on by default for some file types and off for others commonly associated with code and logs. You can also invoke suggestions with Shift + F10. (blogs.windows.com)
By July 8, 2024, mainstream tech outlets reported the rollout to all Windows 11 users, emphasizing that this is a big shift for a tool long considered sacrosanct for raw text. (arstechnica.com)
If you want to practice with or without help, you can flip these settings inside Notepad: open Settings → Spelling, then toggle Spell check and Autocorrect. That menu also exposes per‑file‑type behavior. (pcworld.com)
Why autocorrect changes how we measure skill
Classic typing tests assume a neutral editor—the characters you type are the characters that land. But with autocorrect:
- Some errors never “count” because they’re fixed instantly.
- Your flow can speed up because you trust the editor to catch common slips.
- Your measured accuracy inflates relative to your unaided skill.
That doesn’t mean assisted typing is bad; it just means our metrics should say whether the editor is helping—and by how much.
New metrics for an assisted‑typing world
To keep practice honest yet modern, add a few instrumentation layers to your typing tests and training logs.
1) Autocorrect acceptance rate
- What it is: The share of autocorrected suggestions the typist accepts (explicitly or implicitly) versus rejects/undos.
- Why it matters: A high acceptance rate means the editor is contributing meaningfully to the final text; it also signals where the user relies on correction safety nets.
- How to track: Capture each autocorrect event (word before, word after, accepted vs. undone) and compute Accepted ÷ Total Suggestions.
2) Corrections per 100 words (CP100)
- What it is: All fixes—autocorrected substitutions, manual edits, and suggestion clicks—normalized per 100 words.
- Why it matters: CP100 reveals how much “repair” your text needs regardless of who (you or the editor) did the fixing. A lower CP100 reflects cleaner upstream typing.
- How to track: Increment for every backspace‑borne edit, suggestion pick, or autocorrect replacement. Reset counts between sessions for comparability.
3) Post‑edit latency (PEL)
- What it is: The time between an error and its resolution (e.g., the gap from first incorrect keystroke to correction applied/accepted).
- Why it matters: Shorter PEL suggests better micro‑editing fluency and less momentum loss when mistakes happen.
- How to track: Timestamp keystrokes and edits; for autocorrect, use the event time; for manual fixes, use the key event (e.g., backspace or selection + replacement).
4) Keystrokes saved per 100 words (KS100)
- What it is: Estimated keystrokes you didn’t have to press because autocorrect handled the fix (including skipped backspaces and retyping).
- Why it matters: Quantifies the “assist” in a hardware‑agnostic way, letting you compare devices and settings.
- How to track: For each autocorrect, approximate the avoided backspaces + retyped characters; sum and normalize by 100 words.
5) Backspace rate and undo‑after‑autocorrect
- Why it matters: Some autocorrections are wrong. If undo‑after‑autocorrect spikes, the assist is creating friction or mismatches with a specific vocabulary (e.g., code tokens, jargon).
Experiments you can run today
Below are lightweight studies you can do in Notepad to see how assistive editing shifts your results. Before you start, verify Notepad is up to date via the Microsoft Store and check Spelling settings. Features are enabled by default for some file types, while coding/log‑type files may start with them off. (blogs.windows.com)
A/B: Assist Off vs. Assist On
- Method: Do two 5‑minute passages of comparable difficulty. Session A: Spell check OFF, Autocorrect OFF. Session B: Spell check ON, Autocorrect ON.
- Capture: WPM, visible error rate, CP100, PEL, KS100, and autocorrect acceptance rate.
- Expectation: WPM rises and visible error rate drops in Session B, while CP100 may remain similar if you habitually mistype the same patterns; KS100 will quantify how much the editor helped.
Three‑mode ladder: None → Spell check only → Spell check + Autocorrect
- Method: Run three back‑to‑back 3‑minute tests.
- Capture: Same metrics as above, plus undo‑after‑autocorrect.
- Expectation: Spell check alone may slightly reduce errors via visual cues; adding autocorrect typically boosts WPM further. Track whether post‑edit latency shrinks as you rely on automatic fixes.
Vocabulary stress test (jargon/code)
- Method: Type a domain‑heavy paragraph (e.g., medical acronyms, variable names). Keep your custom dictionary empty for the first run, then add key terms and repeat.
- Capture: Undo‑after‑autocorrect and acceptance rate.
- Expectation: Initial runs show more wrong fixes or red underlines; after dictionary training, acceptance rate should climb and friction fall. This mirrors real‑world desktop writing where personalization improves assistive accuracy. (blogs.windows.com)
Designing “WPM fairness” into your typing tests
To make results comparable across assistance levels, report two speeds:
- Raw WPM: The classic measure from characters/time, independent of editing help.
- Assisted WPM: The displayed speed in the presence of autocorrect.
Then publish an Assist Delta:
- Assist Delta (WPM) = Assisted WPM − Raw WPM
- Assist Share (%) = KS100‑adjusted time savings ÷ total time
This trio lets learners (and leaderboards) compare apples to apples. A user can proudly share a high Assisted WPM while also tracking their Raw WPM improvements as they practice with assistance turned off.
For accuracy, pair visible error rate with CP100. Visible error rate shows what escaped; CP100 shows the total rework burden—including invisible, editor‑handled repairs.
Practical training tips for desktop writers
- Match mode to goal: For exam‑style or coding drills, disable autocorrect (and even spell check) to test unaided precision. For everyday drafting, keep both on to maintain flow.
- Learn the suggestion shortcut: Use Shift + F10 on a misspelled word to apply fixes from the keyboard without breaking rhythm. (blogs.windows.com)
- Customize per file type: Leave assistive tools off for config/log files and code, on for prose and notes—Notepad already respects those defaults, and you can override them. (blogs.windows.com)
- Curate your dictionary: Add names and domain terms so the editor stops fighting you and starts helping.
- Benchmark periodically: Run the A/B test above every few weeks to see whether Raw WPM and CP100 are improving beneath the safety net of autocorrect.
The bigger picture
The desktop has joined smartphones in treating typing as a collaborative act between you and software. With Notepad’s 2024 update, that collaboration is now the default, not the exception. If our training tools acknowledge the assist—measuring acceptance, corrections, saved keystrokes, and latency—we can keep practice fair while embracing a faster, less frustrating writing flow. (arstechnica.com)
