Viral Prediction Isn’t a Vibe — It’s a Supply Chain

TL;DR

We built a Viral Prediction System that predicts how ideas spread using empirical propagation paths (real historical chains), not “synthetic graph vibes.”
The system combines Network Science (bridges vs hubs) with Archetypal Memetics (why an idea resonates).
Under the hood: Scraper → Knowledge Graph → Archetypal Scoring → cadCAD simulation → API.
The user promise is simple: if you’re a broke founder/creator, we help you figure out who to talk to and why the idea will spread—without paid distribution.

Why “Go Viral Advice” Fails

Most virality advice is content astrology: post more, hook harder, copy trends faster.

But virality is a propagation process:

An idea starts in a specific community.
It gets picked up by a bridge (a weak-tie connector) that can move it across clusters.
It mutates to fit the next platform’s format affordances.
It either dies… or cascades.

That shift—treating virality like a supply chain—changes what you build.

The Big Bet: Real Paths > Synthetic Models

A lot of diffusion modeling is clean on paper and soft in production because it starts with a synthetic network and then tries to “seed the top influencers.”

We flipped it:

Start with what actually happened.

Instead of asking, “How could ideas spread?” we ask:

“How did similar ideas spread in the past—and what were the exact bridges?”

That’s what we mean by Empirical Propagation Paths.

Examples of verified cross‑platform diffusion (project notes)

67 Meme: TikTok → Instagram Reels (key spreader: Taylen Kinney)
Kirkification: TikTok (Sept–Oct 2025) → X/Twitter (Charlie Kirk face swaps)
Distraught Morgan Jones: TikTok (Aug 2025) → X/Twitter (Oct 2025)
Nailong Dwerking: TikTok → X/Twitter cross‑platform GIF diffusion

What’s Inside the Knowledge Graph (and Why It Matters)

The milestone isn’t “we made a model.”

The milestone is: the knowledge graph is live—a large, queryable substrate of meme history (memes, events, subcultures, people, sites, cultures) and the relations between them.

Current KG statistics (project notes, Dec 2025):

Entity Type	Count
Memes	33,820
Events	4,012
Subcultures	3,043
People	2,679
Sites	860
Cultures	328
Total Entities	~44,742

Critical architecture note: local vs remote KG

This matters for builders:

Local LightRAG (PostgreSQL) is for user/business knowledge.
Remote poly-scraper-kg is where the scraped meme KG lives.

If you point your app at the wrong store, you’ll get empty results and think the system is broken.

Archetypal Memetics: Fitness Is Psychological, Not Just Statistical

Network position matters. But so does why people share.

We treat memetic fitness as:

Accessibility × Durability

And we score “durability” partly through archetypal balance:

Nature (unknown/chaos/risk)
Culture (order/governance/known)
Individual (hero/adversary/responsibility)

The “sweet spot” is content that is both:

Relatable (personal, concrete)
Resonant (mythic/archetypal)

One‑sided narratives can spike—but also trigger backlash.

System Architecture (Scraper → KG → Scoring → cadCAD → API)

At a high level:

What each stage does

Scraper: ingests and extracts entities/relations from meme sources.
Knowledge Graph: stores entities and relations like originated_in, posted_by, amplified_by.
Workers (query layer): answer questions like “what propagation paths exist for memes in this niche?”
Archetypal scoring: outputs archetypicality, balance score, and ideological risk.
cadCAD simulation: Monte Carlo runs against historical timelines with policies like bridge targeting.
API output: bridge recommendations, propagation maps, reach estimates.

What You Can Do With This (Without a Marketing Budget)

This is built for the person who has more conviction than cash.

The output is not “post 3x/day.” It’s:

Who to talk to first (bridges, not just hubs)
Which platform sequence is realistic for your niche
What mutation the concept needs to cross platforms (format + framing)

Or simply:

It tells you where the doors are—and which key might fit.

What’s Next

Roadmap items (project notes):

Network hybridizer: privilege real chains over synthetic edges
Full archetypal scorer integration into the simulation loop
Frontend visualization (Next.js dashboard)
Backtesting against 100+ known viral events

Sources

Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of “small‑world” networks. Nature. https://doi.org/10.1038/30918
Barabási, A.-L., & Albert, R. (1999). Emergence of scaling in random networks. Science. https://doi.org/10.1126/science.286.5439.509
Granovetter, M. S. (1973). The Strength of Weak Ties. American Journal of Sociology. https://www.journals.uchicago.edu/doi/10.1086/225469
Kempe, D., Kleinberg, J., & Tardos, É. (2003). Maximizing the spread of influence through a social network. KDD. https://dl.acm.org/doi/10.1145/956750.956769
Zannettou, S., et al. (2018). On the origins of memes by means of fringe web communities. IMC. https://conferences.sigcomm.org/imc/2018/papers/imc18-final102.pdf
“Archetypal Memetics: The ULTIMATE Sociological Theory” (YouTube). https://www.youtube.com/watch?v=6JB6YgmeAr4&t=1078s

Call to Action

Want to build or deploy digital workers that can research, reason, and execute (not just chat)?

Explore Poly: https://www.poly186.com
Book a demo: https://www.poly186.com/demo
Documentation: https://docs.poly186.com

Viral Prediction Isn’t a Vibe — It’s a Supply Chain (Memetics + Network Science)