← All publications
Build in Public: Content Machine 1/6
9 years of Zoom meetings. 533 recordings. Thousands of pages of notes in Workflowy and Notion.
Everything was scattered. Inaccessible. Useless.
In one day, I centralized all of it in Obsidian and turned it into a LinkedIn content machine. 15 posts scheduled in 2 languages, published automatically at fixed times, every day.
Here's the full architecture:
🧠 Step 1 — Obsidian as the central brain
A local folder. Markdown files — portable and readable by any AI. Free, fast, 100% offline. No SaaS, no subscription, no vendor lock-in.
📥 Step 2 — Import 10 years of notes
1.8 million characters from Workflowy. 847 MB and 7,837 files from Notion. Direct import, zero data loss.
🎤 Step 3 — 533 Zoom meetings → structured summaries
A Python script downloads transcriptions via the Zoom API. Claude Code summarizes them: decisions, action items, key takeaways. 400 meetings fully processed into Obsidian.
✍️ Step 4 — AI scans the Obsidian vault and writes the posts
Claude Code runs 3 parallel agents. They read 400 meeting summaries. They extract the most impactful insights. Result: 15 posts with real numbers and concrete client cases.
📤 Step 5 — Automatic publication FR + EN
A Python script publishes to LinkedIn via the API. Cron schedules each post: FR at 9:30 AM Paris time, EN 6 hours later. Zero manual intervention.
Total cost: a few dollars in API calls. Time: half a day of setup.
The next 5 posts detail each step so you can replicate the system.
Which step interests you the most?
Everything was scattered. Inaccessible. Useless.
In one day, I centralized all of it in Obsidian and turned it into a LinkedIn content machine. 15 posts scheduled in 2 languages, published automatically at fixed times, every day.
Here's the full architecture:
🧠 Step 1 — Obsidian as the central brain
A local folder. Markdown files — portable and readable by any AI. Free, fast, 100% offline. No SaaS, no subscription, no vendor lock-in.
📥 Step 2 — Import 10 years of notes
1.8 million characters from Workflowy. 847 MB and 7,837 files from Notion. Direct import, zero data loss.
🎤 Step 3 — 533 Zoom meetings → structured summaries
A Python script downloads transcriptions via the Zoom API. Claude Code summarizes them: decisions, action items, key takeaways. 400 meetings fully processed into Obsidian.
✍️ Step 4 — AI scans the Obsidian vault and writes the posts
Claude Code runs 3 parallel agents. They read 400 meeting summaries. They extract the most impactful insights. Result: 15 posts with real numbers and concrete client cases.
📤 Step 5 — Automatic publication FR + EN
A Python script publishes to LinkedIn via the API. Cron schedules each post: FR at 9:30 AM Paris time, EN 6 hours later. Zero manual intervention.
Total cost: a few dollars in API calls. Time: half a day of setup.
The next 5 posts detail each step so you can replicate the system.
Which step interests you the most?