Privacy Notice

The short version

• Muscle Memory is coding practice without AI, and recording how you work is the product: we study how people write code. Your progress and private observed practice statistics are yours now. Stored model tries may be shown separately; community comparisons and research publications are planned.
• You're a random ID in your browser. We don't ask for your name or email.
• Other people never see your results unless you share a result link yourself. Future community comparisons will use aggregates, such as percentiles and averages.
• We may publish insights: statistics and findings about how people code, never your records or your code. Don't want to count towards them? Go stealth in Settings, no questions asked. (Fair warning: stealth works both ways, so you won't see other people's stats, or your own report, either.)
• We don't sell your personal data, and there are no ads.
• What you do here is sent to our servers as you go, and your browser keeps what practice needs (your latest attempt at each challenge). Delete your account in Settings and it's deleted on both.

That's the gist. The rest is the detail, which the law (reasonably) wants written down.

Who we are

Ekku Leivonen, a sole trader running Muscle Memory as rowandreason.com, at Estemcourt 1002, 4-1-20 Nionohama, Otsu, Shiga 520-0801, Japan, is responsible for your data (the "controller", in GDPR-speak). Questions, requests and complaints go to support@rowandreason.com. A human reads it.

What we collect

Your account

• Your profile answers, all optional: experience, what you do, how you learned to code, where you code, languages, how often and for what you use AI, roughly how much of your shipped code an AI wrote, your goal, whether you touch type, an age band, and editor keybindings, and where you gave them.
• Which version of these documents you accepted, and when.
• Your privacy choices, like whether you've gone stealth.

What you do here

• Starting a challenge: which one, where you started it from (the list, quick start, a link, a suggestion), and for a pick from the list, your search and its place in the list. Stars too.
• Pages you view: which page, its search and filters, the page before it, and when the page is visible, hidden, focused or resized. We can tell you switched away, never to what.
• How you work on challenges, keystroke by keystroke: every key you press in the editor with its exact timing and which physical key it was (arrows and shortcuts too), every edit, pastes as the text they add, cursor moves, scrolling, copies and cuts (where in the code, never your clipboard), text you copy from the challenge itself, the type errors and type hints the editor showed you, every run with its output and results, each submission's result and the rules it broke, completing, and your answers to the questions after completing.
• Your setup, coarsely: browser and operating system family, mouse or touch, language, time zone, and what a few keys on your keyboard type (its layout). Never your exact browser version or anything that singles out your device.
• Your profile answers, each time you give one.
• How you found us and what you share: the first page you landed on and when, and if a shared result link brought you, its referral code. When you share a result: which result and how (copy link, X, LinkedIn or your device's share sheet).
• A random ID for your browser, so records from the same person stay together without saying who that person is, and a secret only your browser knows, which proves that ID is yours when it sends records, loads private uploaded statistics or asks us to delete them. Also a separate random referral code, which goes into result links you share.

Keys are heard only inside the editor, never anywhere else on the page or outside it, and we never read your clipboard. Typing rhythm can tell people apart, the way handwriting can. We record it to study how people work, not to identify anyone, and we won't use it for that.

Result links you share

A result link (/r/...) carries the challenge, your language, how its checks went, how many submissions it took and your referral code, and nothing about time or speed. Anyone with the link can read that. Your code goes in only if you tick "Include my code", and then it's in the part of the link after #, which browsers don't send to us, so we never store it. Whoever you send the link to, and wherever you post it, can keep it. We can't take a link back.

When it reaches our servers

• Your browser sends what's above every few seconds, plus the time it arrived. Nothing is sent before you've accepted these terms.
• The country your connection comes from, as our hosting provider places it (Finland, say), stored with your records so we can compare how practice differs between countries. Only the country: never your city, and never your IP address. A VPN or travel can make it wrong, and that's fine.
• The usual server logs: our hosting provider's logs of requests to the service that receives your records, which can include your IP address, browser and timestamps. They're for keeping the lights on and the bots out, and are kept for 7 days. We also use your IP address, briefly and without storing it, to limit how many requests one address can send. Your records don't carry your IP address.

In your browser

• Your practice, progress, stars, profile answers, privacy choices, the ID, its secret and your referral code, kept in your browser's storage (local storage and IndexedDB). Once records are uploaded, the browser keeps only what practice needs: your latest attempt at each challenge.
• One cookie, mm_consent, which holds only the version of these documents you accepted, so pages can let you in. It lasts a year.
• Before you accept, the only thing kept is a note of the page you landed on (and the referral code, if a shared link brought you), in this browser only. It's sent to us after you accept, and never if you don't.

We don't collect your name, email, precise location or contacts, and we don't run analytics services, third-party trackers or ads. The X and LinkedIn buttons are plain links: nothing loads from those sites until you click one. If that changes, this page changes first.

Why we use it

The law asks for a reason (a "legal basis") for everything. Here are ours:

1. Running the product. Your progress, private practice statistics, stars, result cards and suggestions for what to work on next. This is the thing you signed up for. Basis: performing our contract with you (GDPR Art. 6(1)(b)).
2. Making better challenges. Which challenges are too hard or too easy, where people get stuck, and what to write next. Basis: our legitimate interest in improving Muscle Memory (Art. 6(1)(f)).
3. Publishing insights. We think the world could use real evidence on how people solve engineering problems, and how that compares with LLMs. So we may publish statistics and findings drawn from what's recorded. Basis: legitimate interest (Art. 6(1)(f)), with an opt-out in Settings (going stealth). More on how below.
4. Security and abuse. Spotting bots, cheating and attacks. Basis: legitimate interest (Art. 6(1)(f)).
5. When the law makes us. If we're legally required to keep or hand over something. Basis: legal obligation (Art. 6(1)(c)).

Comparisons, and the LLMs

Human comparisons remain unavailable until suitable cohorts and privacy safeguards are ready. Where a verified model receipt exists for the same task revision, the dashboard can show that exact model/version, reasoning setting, tools and observed outcome. A single try is not a stable success probability. Human elapsed time and model latency are different measures.

When they come, comparisons with other people will be aggregates: "faster than 70% of people on this challenge", never a comparison with one named or identifiable person.

Comparisons are a two-way street. If you go stealth, you're left out of everyone else's comparisons, and you don't see other people's stats in return. That includes comparisons with LLMs: stealth means no comparisons at all, and no report on /stats.

Comparing you with LLMs doesn't involve sending your data to them. The models solve the same challenges separately, and we compare the results. No AI model sees what you do here, while you practise or afterwards.

Your private report can summarize your own recorded practice, unless you're in stealth. Separately, we look at records internally: for example, how many people started and solved each challenge, to make better ones. That's internal, so it includes everyone, stealth or not.

Your private report

Your report on /stats is built from what's recorded about your own practice: evidence kept on this device and, automatically when you open it, code-free summaries of your existing uploads from the past year. Your browser proves ownership using its private secret; a participant ID or upload proof alone is not enough. No one else’s activity is returned. The summaries describe observed practice, missing evidence and recorded conditions; they do not measure intelligence, attention or dependence on AI.

The reader adds no new activity collection or permanent summary storage. It processes existing records under the retention periods below. Summaries contain counts, timestamps, recorded runtime/editor conditions, bounded checkpoint labels and optional post-completion answers, not code, raw keys or outputs. Uploaded summaries exist only in request memory and the open page’s memory, while the report is open and for at most five minutes after you leave it; they are not saved in browser storage or included in backup files, and they never change your progress. Reading them is not code recovery or device sync. If a model has tried the same challenges, the report asks for those results too; that request names only the challenge, never you.

Reset clears device progress and the open report, and discards pending responses, without deleting existing uploads. The report doesn't re-read them in that moment; the next time you open it, it reads them again. To remove uploads too, delete your account: that blocks further reads immediately, and existing server erasure completes within 30 days. Logout and privacy changes also clear the report. Stealth means no report at all: /stats reads nothing, neither your history nor anyone's comparisons, and shows only how to change it. Upload delays, missing records and read limits are shown as limitations, never as invented zeroes.

Publishing insights, carefully

What we publish is insights, not data: statistics and findings across many people, like "half of the people who tried this challenge solved it within three submissions", or how typing pace differs with experience. We don't publish your records, your keystrokes, your code or anything else about one person, and we don't hand them to anyone else to study.

• Every figure covers enough people that no one can be picked out of it. Groups too small to hide in are left out or merged, for example a country with only a few people in it.
• Nothing published names you, carries your ID or quotes your code.

Nothing has been published yet. If we ever want to publish or share data about individual attempts, even anonymised, this page changes first and you'll be asked to accept it again.

Published insights aren't about you, and once out they can't be recalled. That's exactly why you can opt out beforehand: go stealth with privacy in Settings. You'll be left out of future insights, including anything built from what you've already done. Stealth doesn't stop recording or uploads: it decides what's published and compared, not what's kept. To stop recording, stop using Muscle Memory, and delete your account if you want what's kept gone too.

Who else sees it

• Cloudflare, our hosting provider: it runs the site, stores your records and erases them when you delete your account, for us and under its standard terms for doing just that.
• Authorities, if the law requires it.
• Everyone, but only the published insights described above, which aren't about any one person.

That's the list. No data brokers, no advertisers.

How long we keep it

• Your account and activity: for as long as you have an account. Delete it and your browser's copy goes at once, our servers stop accepting anything from it straight away, and what they hold is erased within 30 days, old copies included. We keep a note that the random ID was deleted (so nothing can be uploaded under it again) and how many records were erased. Neither says anything about you.
• Server logs: 7 days.
• Published insights stay published. They aren't about you, so there's nothing of yours in them to delete.

Your rights

Depending on where you live (the EU and UK in particular), you can ask us to:

• show you what we have (access), fix it, or delete it;
• give it to you in a portable format;
• stop using it for something, which you can call objecting (going stealth is the one-command version);
• restrict what we do with it while we sort out a disagreement.

Email support@rowandreason.com. We'll reply within a month. You can also complain to your data protection authority. We'd rather you talked to us first, but it's your call.

You can do some of this yourself: delete in Settings deletes your account here and on our servers, export downloads everything your browser holds as a file, and access asks us for a copy of what our servers hold: your browser proves the records are yours and gives you a reference to email us, and we reply with the copy. reset only clears your progress in this browser; what was already sent stays until you delete your account.

Since we don't know who you are, we can only act on a request for records we can tie to you, which is what access does. We keep a note of each request (the random ID, a reference and when). Clearing your browser storage throws away the secret that proves the records are yours, so delete your account first, or export a backup file (it holds the secret, so keep it safe).

Where it's stored

In your browser, and on Cloudflare (our hosting and storage provider), whose storage for your records is in its Asia-Pacific region. Cloudflare runs the site from wherever is nearest to you, and is a US company, so your data may be processed outside your country, including outside the EU/EEA. These are the safeguards:

• We're based in Japan, which the EU and the UK recognise as protecting personal data adequately.
• Cloudflare handles your data under its data processing addendum, which includes the EU's standard contractual clauses (with the UK's addendum). It's also certified under the EU–US Data Privacy Framework and the Global CBPR system.

Ask us and we'll tell you more about any of these.

Uploads are not a backup: they don't give you account recovery or sync between devices. An exported file is how you move to another browser.

Age

Muscle Memory is for people 16 and older. If you're younger, please come back in a bit; the challenges will still be here.

If you're 16 or 17, you get the same protections as everyone, and we've kept them high for that reason: no ads, no third-party trackers, nothing shared with other people, every profile question optional, and stealth and deletion one command away. Nothing here pushes you to give more than you want to. If you give your age band, we treat 16 and 17 separately in anything we publish or compare.

Changes

When this notice changes, the version at the top changes too, and you'll be asked to accept the new one on your next visit. No sneaky edits.

Contact

support@rowandreason.com, or by post to Ekku Leivonen, Estemcourt 1002, 4-1-20 Nionohama, Otsu, Shiga 520-0801, Japan.