← Discern

Privacy Policy

Effective date: August 27, 2026 Last updated: October 4, 2026

Discern judges the quality of an article on your device. By default we never read your browsing history, we do not store full URLs, and we never link analyzed content to your identity. The only exception is an optional, explicitly opt-in “personalization boost” you can grant after install: a one-time local scan that immediately returns the permission (see §7). Our principles: local by default, minimal upload, never tied to who you are.


1. Where your data lives

DataStorageNotes
Your profile (role / sub-domains / level / focus)Local only — chrome.storage.localNever uploaded. Visible and editable to you at any time in the Profile tab
Inferred interests (topic weights, last-read dates)Local only — derived on your device from anonymous per-day topic countsNever uploaded. Shown to you in Profile → “What I’ve noticed” with the evidence behind each one, and removable with one click. No URLs or titles are ever kept — only counts per topic per day
Analysis cache (last ~30 articles)Local onlyExpires after 7 days; stores only title, score, and source hostname — never the full URL
Stats (articles scanned / filtered, etc.)Local onlyPowers the Stats tab; never uploaded
Time ledger (how many articles were filtered per day)Local onlyCounts per day plus one irreversible URL hash per entry, used only to withdraw an entry from the ledger when you mark a verdict “Disagree”. No URL, title or hostname is stored. Day buckets are kept for 730 days, then the oldest are discarded
Capsule position preference (per site)Local onlyStores hostname → position only, no browsing content
Learned fluff phrases (heuristic tuning)Local onlyShort phrases the AI flagged as filler in articles you read; capped in size, used only to tune the local pre-filter, never uploaded
Quota / referral statusLocal + server (anonymous UUID)The server only ever sees a random UUID — never who you are
Pro license status (plan / expiry / bound devices)Server (license key + anonymous UUID)See §6 — payments are processed by Waffo Pancake; Discern never collects payment data

Server-side retention. Everything on the server side lives in our Cloudflare Worker’s key-value store; none of it is exported to a data warehouse or an analytics platform.

RecordRetention
Article text / title sent for AI judgementNot stored — forwarded to the model and discarded in the same request
Anonymous UUID + daily quota counterKept while you keep using Discern; holds only a counter and a first-seen timestamp
Pro license record (license key, plan, expiry, status, up to 3 device UUIDs)While the subscription is active; retained after cancellation so a renewed or reinstated key still works
Referral record / dedup marker30 days
Anonymous error telemetry7 days
Payment webhook dedup ledger7 days
Rate-limit countersAt most 24 hours

2. What leaves your device (and what never does)

Uploaded (only when an AI analysis runs):

  • A sample of the article body plus the title, sent to our Cloudflare Worker, which forwards it to an AI model for judgement.
  • An anonymous UUID for quota counting — not tied to any email, account, or identity.

Never uploaded:

  • ❌ Full URLs (only the hostname is used, and only in the local cache)
  • ❌ Your profile (role / sub-domains, etc. — used only locally for scoring)
  • ❌ Browsing history (sole exception: the opt-in one-time local scan in §7 — results never leave your device), cookies, or form data
  • ❌ Anything that could personally identify you

Not retained after processing: the Worker forwards the AI request, returns the result, and does not persist your article content.


3. Error logs & telemetry

  • Error logs record the hostname only — never the full URL, never article content.
  • Anonymous telemetry (crash / error counts, with no URL and no personal data) is off until you turn it on — the switch lives in your Profile, and nothing is uploaded before you opt in. Uploaded batches are kept on our server for 7 days.

4. AI processing

Article excerpts are forwarded by our Cloudflare Worker to a third-party model (default: OpenRouter / GPT-4o mini) for judgement. We do not use your content to train any model, and requests are not retained after processing.

BYOK mode: if you provide your own OpenRouter API key, analysis requests go directly from your browser to OpenRouter on your own quota — they never pass through our servers.


5. About “shared cache” (a future feature — we will update this policy first)

V1 does not enable any cross-user shared cache. The commitments above describe our full current behavior.

In the future, to make Discern faster and cheaper for everyone, we may introduce an anonymous, content-fingerprinted shared cache: keyed by a hash of the article body, caching only the identity-independent “content quality” judgement (fluff score, verdict, first-hand evidence, content type, etc.). If we do, the commitment will read:

We do not store your URL, do not store your identity, and do not link content to you; to make everyone faster, we may use a content fingerprint to cache the anonymous content-quality result. Personal relevance (match) is always computed from your local profile and never enters the shared cache.

We will update this file and explain it in-product before enabling it.


6. Payments & Pro license

Discern Pro is a paid subscription. All payments are processed by Waffo Pancake — Discern itself never sees, collects, or stores your card number, billing address, or any payment details.

When you purchase Pro:

  • The checkout and payment happen entirely on Waffo Pancake’s side. We do not ask you to create a Discern account — a license key is your credential.
  • Our server stores only what is needed to honor the license: the license key, the plan (monthly / yearly), the expiry date, its status, and up to 3 anonymous device UUIDs the key is activated on. No names, no payment data. The checkout email address is received via Waffo’s order events and used only for subscription support and refunds.
  • Subscription lifecycle (renewal, cancellation, refund) is handled by Waffo Pancake; we receive only the resulting status (active / canceled) via webhook.

Waffo’s privacy practices are governed by their own policy: https://www.waffo.ai/privacy.


7. The optional “personalization boost” (off by default)

To make match scoring smarter from day one, Discern offers one optional boost right after install. It only ever happens if you actively choose it:

  • We never read your browsing history by default. No history permission is requested at install. The capability uses Chrome’s optional permissions: it activates only when you click the grant button and confirm in Chrome’s native prompt.
  • The scan happens on your device, exactly once. If granted, Discern reads your last 30 days of visits locally to detect the marketing topics you actually read (e.g. SEO, CRO), producing a draft profile. Your raw history is never stored, never uploaded, and never leaves your device — only the aggregated topic preferences are kept (and only written into your profile after you confirm).
  • The permission is returned immediately. The moment the scan finishes (or you cancel it), Discern releases the permission — it is held for seconds, then Discern returns to a no-history-access state.
  • Declining costs nothing. Skip it and Discern still works fully — personalization keeps learning silently from the articles Discern analyzes for you, just a little slower.
  • The Chrome Web Store data-use disclosure lists “browsing history” accordingly, as an opt-in-only use.

8. Your controls

  • Turn off telemetry: Profile → “Help improve Discern” switch.
  • Bring Your Own Key (BYOK): use your own OpenRouter key, and analysis requests never go through our quota server.
  • Manage or cancel a Pro subscription: use the “Manage subscription” link in the extension (opens Waffo Pancake’s consumer portal) — cancellation and refunds are handled there, not by us.
  • Uninstall = erased: all local data is deleted when you remove the extension.

9. Children’s Privacy

Discern is not directed to children under 13 (or under 16 in jurisdictions where that is the threshold for digital consent). We do not knowingly collect data from children. If you believe a child has used Discern, please contact us at hello@byzanx.com and we will delete any associated data.


10. Changes to This Policy

We may update this Privacy Policy from time to time. When we do, we will:

  • Update the “Last updated” date at the top
  • Notify users of material changes in-product (or by email, if you have an active Pro subscription)
  • Maintain previous versions available on request (email hello@byzanx.com)

Continued use of the Service after a change indicates your acceptance of the updated policy.


11. Contact

Questions about privacy? Email hello@byzanx.com.

For users in the EU: under GDPR Article 27, our representative for matters relating to GDPR can be reached at the same address.


12. Chrome Web Store Privacy Practices

  • We do not collect personal data except: email + payment details processed by Waffo Pancake when you purchase a Pro subscription (Waffo Pancake is the Merchant of Record; Discern never sees payment card data)
  • We do not sell user data
  • We do not use data for purposes unrelated to the item’s core functionality
  • Data transferred to third parties: (1) anonymous article-text samples to an LLM API for the analysis feature; (2) license key + anonymous device UUID to our backend for Pro verification; (3) payment data to Waffo Pancake for Pro subscription processing
  • Browsing history: never read by default; single opt-in exception is the one-time local personalization scan described in §7, which never uploads history and immediately returns the permission
  • We do not determine creditworthiness or lend

Discern is developed and operated by an independent developer. By using the Service, you agree to the data practices described in this Policy.

Discern
Terms ·Privacy ·Refunds
Support: hello@byzanx.com
© 2026 byzanx.com