A data-driven engine that mines what's already going viral, isolates the hook, and tests it at scale — to launch Teckie, the Discord bot that actually knows your community.
Mine what's already going viral, isolate the hook, and test it at scale — so the data picks the winner, not a guess.
Teckie remembers members, builds their profiles from conversation, summarizes channels, and connects people by what they share. One promise — "it knows your community, so people actually connect" — is what every video sells. Each niche just dresses it in a different story.
We borrow viral-hook fuel from each niche and adapt it to Teckie's promise. Color = niche, all the way through.
Remembers everyone's vibe, then breaks the ice and finds your match.
Never lose what your group covered — summaries + timeline — and find your study twin.
Remembers who plays what and what rank — connects teammates + recaps the drama.
We don't run all three at once. Prove the machine on Dating, then clone it — each round is faster than the last.
Run the full 16-video matrix. Prove the system works.
Swap the hook fuel + story. Same machine.
Repeat. By now the playbook is sharp.
Four hook angles, four core demos, every combination stitched. The grid isolates the hook so the data — not a hunch — names the winner.
Same core, swapped hooks. The hook that lifts performance regardless of core is the real winner.
Same hook, swapped cores. The core that keeps watching once the hook lands is the one to keep.
The opening scene does one job: stop the scroll. It's either existing B-roll or AI-generated video via JSON prompting — whichever best lands the angle.
The first ~1s (rose zone) is the whole game. Win it and the core gets a chance; lose it and the best demo in the world never gets seen.
Four people, four clean handoffs. Each step unblocks the next.
The scraped viral videos are the benchmark — "what good looks like" in the niche, for free.
A video breaks out — ~3–10× your median views — or beats the benchmark on 3-sec hold and watch time.
No breakout, but one hook or core clearly out-performs its row or column. A direction is validated.
The whole batch underperforms baseline → diagnose, don't reprint.
After the 16 go out, the team makes exactly one move — whichever result actually shows up. No branches, no overwhelm.
Make 3–4 more of that exact hook and post them. Why 3–4? It confirms the hook is genuinely good and not just luck — the algorithm sometimes pushes one video at random. Most repeat → real winner, scale it. They flop → it was luck, move on.
Reuse that hook on the next set of cores. Nothing else changes.
Don't remake 16. Ask one question — did people not stop scrolling (hook problem) or stop then leave (core problem)? — fix only that, test just 4.
| Phase | Owner | Status |
|---|---|---|
| 0 · Foundation set — this proposal | Maddy / Bryan | ✅ in progress |
| 1 · Scrape — N8N, 50k+ views | Mike | ◷ backlog · high |
| 2 · Analyze → 4 hook angles | Bryan | 🔒 blocked on 1 |
| 3 · Core videos — secure the 4 core (product/demo) clips | Bryan · Team | 🔒 blocked on 2 |
| 4 · Produce — stitch 4 hooks × 4 cores = 16 | Bryan | 🔒 needs 2 + 3 |
| 5 · Distribute — 4 platforms + Reddit | Team · Dekisora | 🔒 blocked on 4 |
| 6 · Read matrix → one quick-win move | Bryan / Team | 🔒 blocked on 5 |
| → Replicate to Learning, then Gamers | All | ↻ after Round 1 |
All 16 to all 4 platforms (64 uploads), or split the matrix across platforms? Rec: same 16 everywhere — clean cross-platform read.
Are the 4 core videos existing assets, or do they need producing too?
Mike has a ClickUp due date — do we want target dates on Phases 2–6?
Round 1 is hands-on on purpose — we run it manually to truly understand the workflow. Once we're familiar with it, most of it runs itself and we just steer.
Run Round 1 by hand — feel every step, see what actually moves the needle.
Scraping's already automated. Next: the analysis + stitching pipeline.
Stop doing the steps — direct the system instead: review, approve, steer.
Point the same machine at the next niche — Learning, Gamers, beyond.
Mike's Dating-niche scrape unblocks everything downstream, and Dekisora begins Reddit seeding.