47 Steps: How Marina Mogilko turned one recorded interview into content for six platforms

The problem

She counted the steps between hitting stop on a recorder and having something live on every platform. There were 47. “Forty-seven manual steps. Fourteen of them were me personally moving text from one tool to another.” She calls the work “Copy-paste, reformat, upload, schedule.” Her fix was a change in framing: “I stopped treating it like a creative problem and started treating it like an engineering problem.”

The blueprint (as she describes it)

1. Map before automating. She lists every platform in order of strategic importance and writes down each one’s input format. Then she traces every step from source to platform. She circles the steps only she can do, such as creative calls and approvals. She boxes the rest, which are the steps that only move text from A to B. “The boxed steps are your automation targets.”

2. Assign each platform a role. These are the roles on her own map:

   • YouTube is “the discovery engine.”
   • Instagram is “the visual brand layer.”
   • TikTok has “zero dedicated strategy, just cross-posted from the main system.”
   • LinkedIn is for professional authority and “increasingly, AI search visibility.”
   • X carries “podcast clips and occasional tech posts.”
   • The newsletter is her direct email relationship.

   Her core principle is that “one input triggers almost the entire stack.” She still builds separate content for each platform.

3. Transcript into a database. After an interview, the transcript goes into Trint and then lands in the content database, one row per episode. From there, different team members trigger different skills depending on what needs to be made.

4. Skills, not prompts. “A prompt tells an AI what to do once. A skill tells it how we work.” Each skill contains four things:

   • What has worked before, using real posts and clips that performed.
   • Verbatim samples of their best-performing text, for tone and sentence rhythm.
   • Who the audience is.
   • Platform-specific rules.

   To build one, she starts with 5-10 of the best-performing posts on that platform and writes 3-4 sentences on the audience. She then lists the platform’s quirks and adds a banned list. Theirs has about 40 banned words and 12 structural patterns. Each skill is owned by whoever runs that channel: “When LinkedIn output goes stale, the LinkedIn lead updates the skill, not me.” Build time is 2-3 hours the first time and 20 minutes to update.

5. Score clips in parallel. A separate process runs on every transcript. It scores each passage 1-10 against four criteria:

   • A surprising stat.
   • An emotional moment.
   • A counterintuitive claim.
   • A standalone idea that makes sense in 60-90 seconds with no context.

   The process returns the top five with the exact quote, the score, the reason and a suggested text overlay. Claude runs this in about 30 seconds per episode. She says the clip-scoring skill is built on 4 months of their own Instagram data. Anything above a 7 goes to her editors and is edited for Reels and Shorts. It then goes into the same Telegram queue as everything else.

6. One approval queue. After a recording, her team opens Telegram to four drafts: LinkedIn, X, a Telegram update and a newsletter section. She taps approve or reject and leaves a note when she rejects. She says this takes about 10 minutes.

7. Feed analytics back. SVG Command Center is a custom dashboard on Node.js, hosted on Railway. It pulls live data from YouTube, Instagram and Beehiiv into one seven-day rolling view. Three days after every YouTube video goes live, an automation pulls the retention data and writes a post-mortem to the database. She reads it in five minutes. The metric she watches most is the ratio of views to new subscribers.

8. Two Monday audits. One reads the previous week’s data and surfaces what performed, with the number behind it and the element likely responsible. The other scans the week’s workflow logs for manual steps that should not need a human and flags them. She reviews both in about 20 minutes and talks through them with her team on a Wednesday call.

9. Her week. She records a 60-90 minute interview. That evening the repurposing agent fires, clip scoring runs, and everything lands in Telegram before she goes to bed. On newsletter send day the draft already exists, and she edits it for 30 minutes.

10. Her recommended build order. She maps first and builds the database before any automation. The repurposing piece comes before the analytics piece. She writes the voice filter before the platform prompts and sets up one place where everything lands for approval.

Stated results

• Her personal time in the pipeline is about 2.5 hours on a typical one-episode week, including the interview itself. The earlier process had 47 manual steps, but she does not state how much time it took.
• She says starting from one source “cuts my personal time roughly in half.”
• “Every Monday, one or two things come off my plate.”
• The skill discovery audit flagged that TikTok sat outside her analytics layer.
• Gaps she states:
  • Her team overrides about half of what the X skill produces.
  • The voice filter still misses on X about 20% of the time.
  • YouTube comments are read manually and are not fed back into the system.
• Her audience figures are context, not results she attributes to the system. YouTube has 1.55M subscribers, and about 82% of new newsletter subscribers come from YouTube. Instagram has 1.3M, TikTok has 3M, and LinkedIn has almost 70K followers. On LinkedIn she reports 15M+ impressions a year and $80K in direct revenue last year. The newsletter has 21,500 subscribers and a 50% open rate.

Caveats

• This is not a solo setup. She has editors, designers and channel leads who own each skill. She says, “editors are editing, designers are designing.”
• The issue carries a sponsor (HubSpot, tagged #HubspotMediaPartner) and includes a sponsored product mention.
• She writes that “None of it requires you to write code.” She also describes a custom Node.js dashboard. The source does not say who built it.
• The source names Claude only for clip scoring. It does not say which model runs the other skills.
• No before-and-after time figure is given for the 47-step version. No date is given for when the system was built.
• Handles (verified 2026-11-01: linked from her own newsletter header and footer):
  • X: @siliconvalleymm
  • LinkedIn: linkedin.com/in/marinamogilko
  • Instagram: @siliconvalleygirl
  • Threads: @siliconvalleygirl
  • YouTube: @SiliconValleyGirl

FAQ

Does Marina Mogilko run this system alone?

No. Channel leads own each skill, and editors and designers do the production work. She says the system removes the bottleneck of her touching everything.

What goes into one of her skills?

Each skill holds four things: what has worked before, verbatim samples of their best-performing text, who the audience is, and platform-specific rules. Their banned list has about 40 banned words and 12 structural patterns.

How much of her own time does the pipeline take?

About 2.5 hours on a typical one-episode week, including the interview itself. Approving the four Telegram drafts takes about 10 minutes, and the newsletter edit takes 30 minutes.

What still does not work?

She says the X skill comes out too structured and gets overridden about half the time. The voice filter misses on X about 20% of the time. She reads YouTube comments manually.

Sources

• https://siliconvalleygirl.beehiiv.com/p/my-content-system (primary: first-person issue of her Future Proof newsletter, Jul 9, 2026)

---

Want agents like these working your distribution? Get It Built. Get new teardowns weekly: subscribe (double opt-in).