Atomizing Content
YAML--- name: atomizing-content description: Extracts standalone ideas from long-form content (podcasts, webinars, videos, transcripts) and rewrites each as native, platform-specific posts for X, LinkedIn, Instagram, and newsletters. Use when repurposing a single piece of long content into a ranked, ready-to-schedule multi-platform content package. --- # Atomizing Content
Given a transcript or long-form source, produce a ranked set of platform-native posts:
- Read the full source.
- Extract 15–30 candidate "atoms" (standalone ideas).
- Score each atom for standalone strength.
- Keep the top ones (typically top 50%).
- Rewrite each surviving atom natively for its target platform(s).
- Output a ranked, labeled package ready to schedule.
No preamble — go straight into extraction once source is provided.
Progress:
- Step 1: Ingest and chunk the source
- Step 2: Extract candidate atoms
- Step 3: Score and rank atoms
- Step 4: Assign platforms per atom
- Step 5: Rewrite natively per platform
- Step 6: Assemble final ranked package
Step 1: Ingest and chunk the source
Break the transcript into semantic chunks (by topic shift, not fixed time/length). Note timestamps or section markers if available — useful for later clipping.
Step 2: Extract candidate atoms
An atom is a single idea that:
- Makes sense without the surrounding context
- Has a clear point, claim, story, or insight
- Could be understood by someone who never saw the source
Pull atoms liberally at this stage — aim for quantity. Typical sources yield 15–30 raw candidates. Capture each atom as: core idea (1 sentence), supporting detail/quote, source location.
Step 3: Score and rank atoms
Score each atom 1–10 on:
- Standalone strength — does it need zero context?
- Specificity — concrete claim, number, example, or story (not generic advice)
- Tension or novelty — surprising, contrarian, or emotionally resonant
- Reusability — clean hook potential, quotable, visual, or demonstrable
Average the four into a composite score. Cut the bottom half unless the piece is short on material. Rank survivors highest to lowest — this ranking determines publish order.
Step 4: Assign platforms per atom
Match atom type to platform:
- Narrative/process/numbered insight → X/Twitter thread
- Professional lesson, career/business angle, framework → LinkedIn post
- Visual, demonstrable, or emotional moment → Reel/Short script
- Deeper context or several related atoms combined → Newsletter section
High-scoring atoms with cross-platform appeal get rewritten for multiple destinations.
Step 5: Rewrite natively per platform
Never copy-paste the same text across platforms. Each rewrite must match the platform's native reading pattern:
- X thread: Hook tweet (curiosity/claim) → numbered body tweets, one idea each, short lines → closer with takeaway or CTA.
- LinkedIn: Hook line (no bait-y clickbait) → whitespace-heavy paragraphs → personal/professional framing → single clear takeaway → soft CTA (question or opinion ask).
- Reel/Short script: On-screen hook line (first 1.5 sec) → spoken script with beat-by-beat pacing notes → visual/b-roll suggestions → caption text separate from spoken line.
- Newsletter: Section headline → 2–4 short paragraphs → pull quote or stat callout → link back to full source.
Step 6: Assemble final ranked package
Output format:
#1 [Score: 9.2] — "Atom summary"
Platform: LinkedIn
Source: [12:34]
---
[Full rewritten post]
---
#2 [Score: 8.7] — "Atom summary"
Platform: X Thread
...
Order by composite score, not by source chronology.
Example 1: Input: 45-minute podcast transcript about a founder's failed first startup. Output:
- Atom: "I ignored a warning from my co-founder for 6 months and it cost us $200K" → scored 9.1 → rewritten as X thread hook + 8 tweets walking through the mistake, timeline, and lesson.
- Atom: "The real reason most startups fail isn't the product, it's founder conflict" → scored 8.4 → rewritten as LinkedIn post with personal framing and discussion CTA.
- Atom: Founder re-enacting the moment he found out the company was out of runway → scored 8.9 → rewritten as Reel script with on-screen hook "I had 11 days of runway left and didn't know it."
Example 2: Input: 90-minute webinar on email marketing tactics. Output: 22 extracted atoms → 11 survive scoring → final package of 6 LinkedIn posts, 3 X threads, 2 Reel scripts, 1 newsletter roundup combining the remaining lower-scored atoms as quick tips.
- Extract before scoring — don't filter ideas while reading, it kills volume.
- One atom = one idea. If a candidate needs two "ands" to explain, split it.
- Preserve the speaker's specific language, numbers, and phrasing — specificity sells, paraphrasing dilutes.
- Rank ruthlessly; a ranked list of 10 strong atoms beats an unranked list of 25 mediocre ones.
- Flag atoms suited to multiple platforms rather than forcing a single destination.
- Note source timestamps so video/audio clips can be pulled later for Reels/Shorts.
- Don't summarize the whole episode — summaries aren't atoms, they lack standalone punch.
- Don't reuse identical copy across platforms; a thread rewritten as a LinkedIn post by just removing numbers still reads like a thread.
- Don't keep atoms that rely on "as I mentioned earlier" or unexplained inside references.
- Don't score generously — a content package is only as good as its weakest included atom. Cut more than feels comfortable.
- Don't bury the strongest atom — lead the package with the highest-scoring piece, not the first chronological one.