@cryptotaxi247 / SquadScope-1 / commits / 56ad3a9

docs: update methodology and privacy pages for podcast launch (#392)

Addresses #340. Updates methodology page with Signal Check podcast production details and privacy policy with TTS provider, staging storage, and platform disclosures. Resolves Copilot review comments: - Narrowed listener analytics claim (platform may provide aggregate stats) - Fixed AI voice disclosure (synthetic personas, provider-side training, intro+outro) - Verified segment names match canonical editorial style guide Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Juan Manuel Servera committed Jun 12, 2026 at 03:53 UTC 56ad3a91b95e0c3fe822546f7a597e4398957608
2 files changed +58 -1
content/methodology/_index.md
+29
@@ -47,6 +47,35 @@ For topic channels, `scripts/score_repos.py` can compute a 0-100 relevance score
47
48 These caveats are part of the analysis, not a disclaimer that makes the output neutral. The AI-assisted write-up can still over-weight vivid examples, repeat gaps in source coverage, or describe a project more confidently than the evidence supports. Reader reports are welcome when a claim, framing, source choice, or safety implication needs review.
49
50 +## Signal Check podcast
51 +
52 +Claracle publishes a companion podcast, **Signal Check**, derived from the weekly written article. Signal Check is an 8–10 minute two-host conversational episode that highlights the week's signal picks, noise calls, gaps, and predictions. It is distributed through an external podcast platform (currently Spotify); the Claracle website links to the external podcast page but does not host audio.
53 +
54 +### How Signal Check is produced
55 +
56 +1. **Source:** every podcast episode is generated from the corresponding published weekly article and its underlying claim ledger. No additional sources are introduced for the podcast that were not already cited in the article.
57 +2. **Script generation:** an AI-assisted pipeline converts the article into a structured script following a locked segment order (Cold Open, The Signal, The Noise Check, The Gap, Receipts Round, Week Ahead, Outro). The script is constrained to substantive claims that appear in the source article's claim ledger.
58 +3. **Claim ledger:** every factual claim in the script (star counts, growth rates, project characterizations) must trace back to a claim in the published article or the underlying data artifacts. Claims that cannot be traced are removed before synthesis.
59 +4. **Human review:** scripts and publishing packets are reviewed by a human operator before any non-dry-run audio synthesis or public distribution.
60 +5. **Synthesis:** after approval, the script is converted to audio using AI text-to-speech. The voices are synthetic; no human narration is used or implied.
61 +6. **Publishing:** episodes are published manually to the podcast platform with AI-generated-voice disclosure, source article link, corrections link, and show notes.
62 +
63 +### What the podcast does not change
64 +
65 +- The weekly written article remains the primary Claracle product and is published independently. Podcast generation cannot delay, gate, or modify the article publishing pipeline.
66 +- The podcast does not add new sources, new analysis, or new claims beyond what appears in the published article.
67 +- If the podcast generation fails for any reason, the written article is still published on its normal schedule.
68 +
69 +### Podcast-specific bias and limitations
70 +
71 +- **Performative framing:** conversational audio can make claims sound more authoritative or entertaining than their textual equivalents; listeners should consult the source article and claim ledger for precise evidence.
72 +- **Compression loss:** a 10-minute audio format cannot include every nuance from the written article; some caveats may be abbreviated.
73 +- **Synthetic voice limitations:** AI-generated voices may mispronounce names, flatten tone, or introduce cadence artifacts that affect comprehension.
74 +
75 +### Corrections for podcast content
76 +
77 +Podcast corrections follow the same process as article corrections (see below). If an error originates in the source article, it is corrected there and the podcast episode show notes are updated. If an error is introduced during script generation (for example, a claim that does not trace to the ledger), it is corrected in the script and, if feasible, the audio is regenerated or an errata note is added to the show notes.
78 +
79 ## Corrections and reader reports
80
81 Weekly articles include a footer link to report a correction, source concern, or safety concern through GitHub Issues. The link prefills the article URL and week so reports can be triaged against the exact published page.
content/privacy/_index.md
+29 -1
@@ -57,6 +57,32 @@ Under GDPR Articles 15–22, you may have rights to access, rectify, erase, rest
57
58 Because Claracle holds no first-party visitor data, most practical data-subject requests need to be exercised directly with the third party that controls the data: [GitHub privacy requests](https://docs.github.com/en/site-policy/privacy-policies/github-privacy-statement) for hosting logs, and [Google privacy controls](https://policies.google.com/privacy#infocontrols) or [Google's privacy request tools](https://support.google.com/policies/troubleshooter/7575787) for GA4 or Fonts data. You can still [open a SquadScope issue](https://github.com/jmservera/SquadScope/issues/new) if you need help identifying the right route.
59
60 +## Signal Check podcast
61 +
62 +Claracle publishes a companion podcast, **Signal Check**, on an external podcast platform (currently Spotify). The podcast is produced using AI text-to-speech; no human voice recordings are collected or stored.
63 +
64 +### What data the podcast pipeline processes
65 +
66 +- **Public article text and claim ledger:** the podcast script is generated from the published Claracle article and its structured claim data. These are public materials already available on the website.
67 +- **Text-to-speech provider:** the generated script is sent to a third-party TTS provider (currently Azure AI Speech) for audio synthesis. The provider receives the script text. No personal data about readers or listeners is included in TTS requests. Azure AI Speech processing is governed by the [Microsoft Privacy Statement](https://privacy.microsoft.com/privacystatement) and the service's data handling terms.
68 +- **Temporary staging storage:** generated audio files are temporarily stored in Azure Blob Storage for operator review before publishing. Access is restricted to project maintainers; files are retained only until publishing is complete and then deleted per a configured retention policy.
69 +- **Podcast platform:** published episodes are hosted on Spotify (or another platform if changed). When you listen through the platform, that platform's privacy policy applies — Claracle does not control or access listener-identifying personal data from the podcast platform. The platform may provide aggregated, non-identifying listen statistics (such as total play counts) to the podcast owner; Claracle does not use any platform-provided analytics for tracking, profiling, or advertising.
70 +
71 +### What we do not do
72 +
73 +- We do not collect or store listener-identifying personal data (email addresses, listening habits, or individual profiles) through the podcast.
74 +- We do not use listener data for advertising or share it with third parties.
75 +- We do not embed tracking pixels, analytics beacons, or advertising identifiers in podcast audio.
76 +- We do not run dynamic ad insertion.
77 +
78 +### AI-generated voice disclosure
79 +
80 +All podcast audio is synthesized using artificial intelligence. The hosts are AI-generated synthetic personas produced by a text-to-speech model; no human narration is used or implied. No voice recordings are collected or stored by Claracle, and the resulting voices do not represent or impersonate any real individual. TTS models may have been trained on licensed voice talent recordings by the provider (see the provider's terms for details). AI-generated voice disclosure is included in each episode's intro, outro, show notes, and platform description.
81 +
82 +### Future changes
83 +
84 +If Claracle adds podcast analytics, listener support/donations, sponsorships, premium content, or changes the TTS provider or hosting platform, this privacy policy will be updated before those features launch.
85 +
86 ## How to withdraw analytics consent
87
88 Use the **Manage cookies** button in the footer to reopen cookie preferences, turn off Analytics, and save. You can also delete Claracle cookies in your browser settings. Withdrawal stops future GA4 events from this browser; it does not automatically delete data already controlled by Google.
@@ -77,7 +103,9 @@ GitHub and Google may process data in countries outside your own. GA4 data may b
103
104 ## Changes to this policy
105
80 -Last updated: 2026-05-25. Changes are announced through the git history of this page in the public SquadScope repository, so you can review what changed and when.
106 +Last updated: 2026-06-12. Changes are announced through the git history of this page in the public SquadScope repository, so you can review what changed and when.
107 +
108 +**2026-06-12:** Added Signal Check podcast section covering TTS provider, staging storage, and platform disclosures.
109
110 ## Contact
111