Version 1.9.1.7 · Last updated 24 June 2026 · Contact AOEPlayingFild@protonmail.com

The short version

>=PlayingFild classifies pages as productive, neutral or unproductive, and can close tabs you have told it are unproductive. To do that well it reads the page title and visible text on most sites, and classifies them on your device.

It does not read page content on banking, login, payment, webmail, healthcare or AI chat surfaces. Those get hostname-only handling, a Neutral classification, and no content scraping at all.

You sign in with a verified email before it classifies pages, closes tabs or shows feedback cards. In Local mode your classifier data stays on your device and nothing is sent to our servers. In Standard mode a limited set of anonymised signals is sent, described in full below.

We do not sell, trade, rent or transfer user data, browsing history or personal profiles to advertising networks, data brokers or anyone else.

How we decide not to read a page

Three independent layers run before any page content is read. If any one of them triggers, content scraping is skipped entirely.

• Host blocklist. Known sensitive categories: banks, webmail, healthcare portals, password managers, payment processors, government services, listed AI tools, adult content.
• URL path rules. Auth and payment paths such as /login, /checkout, /oauth, /billing.
• On-page detection. If a page looks like a bare login form, a payment form or an AI chat interface, we skip reading it even when the hostname is not on any list. This covers unlisted banks, regional services and AI tools we have never heard of.

When any layer triggers: page body text is not read, no feedback card is shown, the tab is classified Neutral, and nothing is sent to our servers for that visit. Time on the site still appears in your own local stats.

A limitation worth stating plainly: no static list can cover every bank or AI tool in the world. The on-page detection layer exists to catch what the lists miss, but it is not a guarantee against every possible page layout.

The four layers of data

Layer 1: processed on your device

Never leaves your device, in any mode: full page URLs, titles, headings and visible text used for classification. Banking source and target lists. Timer lengths and startup tab URLs. Daily typing-speed charts. Raw click coordinates, scroll depth, keystroke intervals and per-tab dwell time.

Stored locally and synced in Standard mode: your personal keyword weights, hostname bias and path scores, class counts, and an internal trust score that weights your feedback against other users.

Layer 2: Standard mode telemetry

Only active if you choose Standard mode. What gets sent depends on the site.

Most sites: hostname only. example.com, not example.com/some/deep/path.

A curated list of public content sites: full path. YouTube videos, Reddit threads, Wikipedia articles, public GitHub repositories, Stack Overflow questions, Medium articles and similar. On these, the path identifies public shareable content rather than personal state, and it materially improves classification.

Personal paths on those sites: hostname only anyway. Settings, account, billing, security, notifications, private messages and user profiles get stripped back to the hostname even on allowlisted hosts.

A note on code hosting. GitHub URLs contain user, organisation and repository names. For public repos that is already public. For private or enterprise repos, the URL reveals org and project names even though we never see the code. If your repository names should stay private, use Local mode. Gists are excluded from telemetry entirely, because people put personal snippets and config files in them.

Never sent in Layer 2: full page body text, your email or display name, raw mouse coordinates, keystroke content, URLs containing email addresses, OAuth callbacks, session tokens, API keys, or anything at all from excluded or sensitive pages. Tracking parameters (utm, fbclid, gclid) are stripped before transit.

Layer 3: account sync

When you sign in, your personal scores and learned vocabulary sync across your devices. Stored against your verified email and protected by database row-level security, so no other user can read your row.

Layer 3 does not sync your banking source and target lists, your startup tab URLs, or your full timer configuration. Only an allowlisted set of window preferences.

Layer 4: anonymous product analytics

A small number of anonymous events about how people use the extension itself: install, tutorial started, tutorial step reached, sign-in reached, first focus session, uninstalled. Ten event names in total, on a fixed allowlist.

These contain no URLs, page titles, hostnames, tab IDs, personal identifiers or free text of any kind. Every property value is bounded and checked against a per-event allowlist before sending, so free text is physically prevented from entering the pipeline. The anonymous install ID is never joined to your email, account ID or telemetry ID, and the code that sends these events has no access to your authenticated identity.

Stored with PostHog Cloud EU, deleted after 12 months. Turn it off any time in Profile → Advanced Settings.

This layer exists so we can see that, say, most people reach tutorial step five but far fewer finish sign-in, without ever looking at an individual’s browsing.

Why we measure engagement

In Standard mode the shared classifier learns from user feedback, which makes it vulnerable to poisoning. Someone scripting fake traffic could flag productive sites as unproductive for everyone.

To prevent that, the extension derives an engagement score for each page visit, on your device, from active focus time, whether the tab was interacted with at all, scroll depth in buckets, and whether there was more than one interaction.

What is sent is a single decimal between 0 and 1, paired with the sanitised site identifier. Raw click coordinates, mouse trajectories, the characters you type and keystroke timing are never recorded, transmitted or stored, in any mode. If an account consistently submits near-zero scores across many domains it is statistically likely a bot, and its feedback is excluded from the shared model.

How this is funded

>=PlayingFild is free. A paid tier with fully local processing may be offered in future; it does not exist today, and this policy will be updated before it launches.

We will never sell your data to advertising or retargeting networks, surveillance companies, data brokers, insurance underwriters, credit scoring services, employers or hiring platforms, or government agencies except in response to a valid legal order. If we ever consider a category not on that list, we will update this policy, email every signed-in user, and require fresh explicit opt-in.

Browser permissions

• tabs, scripting, webNavigation for tab management, classification and in-page UI
• storage, unlimitedStorage for local classifier data, settings and daily stats
• alarms, idle for timers, break sessions and idle-aware time tracking
• notifications for optional alerts when tabs close
• all sites because content scripts must run on the pages you visit to classify them

Password fields and payment inputs are never included in scraped text.

Age requirement

Not for users under 13, under UK GDPR, EU GDPR and COPPA. You confirm your age at signup. We store the timestamp of that confirmation and nothing else. We never ask for your birthdate.

Your rights

Email AOEPlayingFild@protonmail.com to exercise any of these. Deletion requests are processed within 30 days.

• Know what we hold. This document is the answer. Specific questions welcome.
• Delete your account data (Layer 3).
• Delete your telemetry contributions (Layer 2). Send your telemetry ID and we remove every matching row.
• Opt out of anonymous analytics (Layer 4) in Advanced Settings, or send your anonymous ID to delete past events.

One thing we cannot undo, stated plainly: once individual contributions have been mathematically fused into multi-user averages and global model weights, they can no longer be isolated or attributed to anyone. Deletion removes your rows; it does not unwind aggregate statistics that were already computed. Those aggregates are anonymous and are retained permanently to keep the classifier stable.

If you live in the EU, UK, California or Australia, you also hold the rights granted by GDPR, UK GDPR, CCPA and the Privacy Act 1988.

Retention

• Layer 1 stays on your device until you uninstall or clear browser storage.
• Layer 2 rows are kept while your account is active, and purged on request.
• Layer 3 is kept while your account exists, removed within 30 days of a deletion request.
• Layer 4 events are deleted automatically after 12 months.

Security

Everything in transit is encrypted over HTTPS/TLS. Account data is isolated with database row-level security. Authentication uses Supabase Auth with industry-standard hashed credentials. Analytics use project-scoped keys with no shared secrets in the client bundle.

Changes to this policy

If this changes we will update the date above, email every signed-in user, and show a notice in the extension the next time you open the dashboard. Material changes require fresh explicit opt-in.

Chrome Web Store limited use disclosure

>=PlayingFild’s use and transfer of information received from Google APIs adheres to the Chrome Web Store User Data Policy, including the Limited Use requirements. The extension uses only the permissions required for its stated features. We do not sell, trade, lease or rent user data, browsing metrics or telemetry to third parties, advertising networks or data brokers under any circumstances. User data is not used for remote code execution, for tracking behaviour across unrelated platforms, or for alternative monetisation pipelines.

---

>=PlayingFild is built by a solo developer in Australia. Privacy questions, requests and concerns: AOEPlayingFild@protonmail.com