# Source-grounded audience and content validation

## Evidence boundary

Before synthesis, separate material into three labelled layers:

1. **Source** — explicitly present in supplied audio, video, transcript, workbook, or named expert's text.
2. **Margarita / Cosmos Luminosity material** — her preferences, homework, vision, products, and approved philosophy.
3. **Strategic adaptation** — the agent's proposed application, experiment, metric, funnel, or technical solution.

If Margarita requests “только то, что было в материалах”:

- return only layer 1 and explicitly requested parts of layer 2;
- do not introduce product routes, subscriptions, automation, MVPs, metrics, or platform features unless directly present in the named source set;
- English orientation may be applied only as a change of research language/market unless the sources contain a fuller international-market method;
- state plainly when the source does not contain a dedicated English-market playbook.

## Correct sequence before website or automation

1. Choose the people Margarita genuinely wants to work with.
2. Mine current public voice-of-customer language.
3. Interview representatives without pitching or leading.
4. Identify repeated situation, trigger, desired result, attempted alternatives, and purchase behaviour.
5. Create one minimum paid test.
6. Sell and deliver it manually.
7. Collect completion, outcome, objection, and continuation evidence.
8. Revise or repeat.
9. Design the durable site path and automate the repeated process only after validation.

## Audience evidence standard

Do not infer wealth, gender, profession, or readiness to pay from aesthetics or location. Prefer behavioural evidence:

- independent purchase authority;
- prior payments for adjacent support, education, retreats, therapy, coaching, creative or energetic practices;
- an active search triggered by a real transition;
- dissatisfaction with attempted alternatives;
- ability and intention to apply the result.

Save exact English quotes before translation. Every public quote used as evidence needs a URL and date where available.

## Katya Ro Reels acquisition method — source notes

From the Universe transcript, Katya Ro's stated method is:

1. Search Instagram Reels using topic hashtags, in Russian or English.
2. Find Reels with around one million or more views.
3. Save them into a separate collection.
4. Copy the **form**, not the creator's meaning or text.
5. Put an original story, value, and audience-specific meaning into that form.
6. Use very short (~5-second) archive footage: face, walk, weather, nature, etc.
7. Use no or minimal editing.
8. Publish one Reel daily.
9. Continue the practice for three months.
10. Combine reach Reels with expert Reels/carousels, which may get fewer views but let people try the knowledge.
11. Package deeper information into small guides.
12. Speak specifically to the desired audience; follower quantity turns into quality only under that condition.

Related source details:

- Katya says her first million-view Reel appeared about one month into the daily practice.
- Her example Reel reached about 1.7M views and brought thousands of followers; this is a case, not a guarantee.
- Her homework asks for 10 topic Reels that can be recreated in about 10 minutes, without editing or preparation, with one idea and optionally no face.
- Popularity alone is not audience fit. A viral topic aimed at a different lifestyle can bring the wrong followers.

## Finding and verifying concrete Instagram Reels

When Margarita asks for actual popular Reels, return direct `instagram.com/reel/...` links, not creator accounts, hashtag pages, generic examples, or invented hooks.

A practical public-discovery workflow:

1. Open Instagram's public topic page, e.g. `https://www.instagram.com/popular/<topic>/` or an Explore-tag page.
2. Use the rendered page rather than relying on search-engine indexing; public topic pages can expose Reel cards, captions, authors, direct URLs, and view counts without login.
3. Extract direct Reel links and visible card text from the DOM. A useful browser-console expression is:
   ```js
   [...document.querySelectorAll('a')]
     .filter(a => a.href.includes('/reel/'))
     .map(a => ({
       url: a.href.split('?')[0],
       text: (a.innerText || '').trim()
     }))
   ```
4. Search several audience situations separately (`multipassionate`, `career-change`, `starting-over`, relocation, divorce, self-trust) rather than trusting one broad hashtag.
5. Rank only after filtering for relevance. A high-view Reel with the wrong job-to-be-done is not a useful model.
6. Record the view count exactly as publicly displayed and the date observed. Never infer hidden likes, date, audience demographics, or historical peak metrics.
7. If the creator's normal reach is visible, compare against it. If it is not, say that only raw public views were verified.
8. Describe the result as “the most popular among the relevant Reels found and publicly verified,” not “the ten most popular on Instagram.” Instagram does not expose a complete global ranking.
9. Reopen or spot-check every final URL before delivery when access permits.

If a topic page is unavailable, vary the topic phrase or use an adjacent public topic page. Capture the successful retrieval pattern, not a permanent claim that a search engine or Instagram is broken.

## Applying the method without plagiarism

For every saved Reel record:

- direct URL and date observed;
- author and visible view count;
- creator's normal view range when verifiable;
- first frame;
- exact on-screen hook or exact visible caption, clearly distinguishing the two;
- duration when verifiable;
- visual action;
- emotional function;
- caption structure;
- CTA;
- who identifies in comments, only if comments were actually inspected;
- why someone would share it;
- the original Cosmos Luminosity meaning that could inhabit the structure.

Do not copy wording, footage, story, or identity. Reuse only abstract grammar such as:

- `POV: you realize…`
- `Maybe you’re not X. Maybe…`
- `The hardest part of X isn’t…`
- `You don’t need X. You need…`

When reporting a Reel library, explicitly separate:

- **verified facts:** URL, author, visible views, visible caption;
- **interpretation:** why it fits the segment and what emotional job it performs;
- **CL adaptation:** an original proposed hook or meaning.

Never present a caption excerpt as verified on-screen text unless the video itself was inspected.

## Validation reporting

Keep these metrics distinct:

- reach / views;
- profile visits;
- relevant followers;
- comments showing self-recognition;
- saves / shares;
- guide or trial uptake;
- conversations;
- paid purchases;
- product completion and result.

Do not report reach as product validation or follower growth as willingness to pay.