frameworkyoutube-viral-videos · v11099:cursor-joseph-os2026-09-14served from databaseAll documents

FrameWork Youtube Viral Videos

For: operators and agents packaging long-form YouTube (and podcast-on-YouTube). Slug: youtube-viral-videos Page: https://projects.jbnx.io/framework/youtube-viral-videos Source: Paddy Galloway × Open Residency, 2026-09-14, Z2uoA3bhJT0 (2:45:08). Auto-captions; names below are corrected from ASR.

How to pull the next transcript: FrameWork Youtube Transcripts.

This is a click-and-watch playbook. YouTube is not a hosting dump and not a "serve me Reels" feed. The viewer still chooses to click, then must stay.

1. The game in one screen

Layer	Question	Fail mode
Idea	How far into the TAM can this travel?	Great execution on a tiny idea (ceiling)
Package	Will a stranger click in one second?	Vimeo-hosting: film → upload → pray
Watch	Do they stay, and leave satisfied?	Promise in the title, bait in the first 10s
Repeat	Can we do it again in this niche?	One viral, then random

Algorithm inputs he trusts: people clicking, people watching, people being satisfied. Primary metrics in Studio: views (comprehensive) + retention curve (where they leave). Do not manage to CTR% or AVD as the scoreboard — those move with audience mix (lemonade stand at home vs Times Square).

Time split (100 hours on a video): 10–20 hours on packaging. Packaging is ~50% of the outcome, not 50% of the clock.

2. Idea: ceiling, floor, TAM

Three circles, looked at together:

All viewers you could reach — niche / TAM. Knitting < MrBeast. Bigger TAM = more competition and more upside.
Ceiling the idea imposes — a technical "how I lined my wildlife pond" dies at ~10k; "I built a wildlife pond in my backyard" did 25–30M. Same pond.
Execution — only fills the circle the idea opened.

Floor: if you posted a trash video titled "video 1" with your face, how many still watch? That is the dedicated core. Everything above the floor is execution.

Astrophotography proof: 1–2k views on telescope reviews → "Milky Way in 10 min vs 1 hour vs 100 hours" (format stolen from a haircut video) → 2–3M.

Hobbyist vs operator: if the goal is the biggest channel in the niche, every idea needs evidence it can travel. "I thought it was interesting" is not a strategy.

Skill ∩ demand: do not chase Minecraft because it is hot if that is not your skill. Intersection of what you are good at and what YouTube is rewarding.

Study 100-sub outliers, not only MrBeast. A 100-sub channel that hits 3k (30× average) is a VC seeing $10k → $30k MRR. Tools: 1of10, Viewstats, vidIQ — the logo does not matter; you need channel average vs this video.

Rough sanity (vanity, not algo): ~20% of subs as "normal" views. Overperformance vs that is the tell. Subs in 2026 are a scoreboard, not the ranking lever.

3. CCN — Core, Casual, New

Every upload should be watchable by:

Core — ride-or-die (10k true fans). At zero you have none.
Casual — 1-in-3 / 1-in-4. Convert them toward core.
New — remaining TAM you have not reached this upload.

Best videos hit all three. Core-only = studio tour / "official reveal of my new YouTube studio" (Jesser: 400k). Core+casual = familiar challenge, no new TAM (~1M). Core+casual+new = "5 NBA games in 50 hours" (14M) — effort signal, one-second glance, basketball TAM.

Traps:

Listening to comments ("faster horse") → more of the old thing that no longer grows.
Chasing only new: no through-line, whatever is viral this week.
Assuming overlap that does not exist. Red Bull mixing parkour + F1 + soccer + Eileen Gu is staff taste, not audience overlap. Call of Duty ∩ Battlefield ≈ 20%. Bodybuilding ≠ yoga. Start narrow, expand outward (Forest Galante: extinct animals → zoos / big cats). You cannot start broad and later niche down.

80% strategy rule (not Pareto): 80% of videos should have ~80% topic overlap with each other (could the same 80% of viewers watch all of them?). Experiment in the other 20% (1 in 4–5). Geniuses (Casey Neistat) can ignore this; operators cannot.

4. Hallmarks of 1M / 10M / 100M videos

CCN fit.
Glance test / log line. One second, one sentence. If it takes 30–40s to pitch, it is not an idea, it is a thought. Complicated titles kill entrepreneurs.
Universal appeal — interesting to the biggest YouTuber in the niche and your mother. Kids' movies that still work as an adult. Never assume the new viewer saw last week. Do not kiss "Katie" without sewing "she's my girlfriend" into the line. Never "as we discussed last week."
A question the click answers. "5 NBA games in 50 hours" → did he, which games, how. Click-to-unpause. If they are about to do something important, is this strong enough to watch first?

5. Where ideas come from (internal / external / innovation)

Put all three at the top of a funnel, then eliminate.

Source	Meaning	Move
Internal	What already worked for you	Repeat / twist / double down fast — others will copy in 2026
External	Outliers on other channels in-niche and adjacent	≥3–4× channel average. Golf → fishing (same demo, equipment hobby). Steal title, thumbnail, format — rarely "the edit"
Innovation	Novel combo	~20% of a small channel's slate. Paddy: Hollywood video-essay × business whiteboard → YouTube-channel essays (2016)

Start from scratch: dump every 2-year outlier in-niche + adjacent, write your spin on each. 5k vs 1k average is still data.

Frankenstein (Noah Kagan, 2021): Hormozi-style "I asked a billionaire…" title + Canadian "can I cook in your kitchen?" door thumbnail + Daniel Mark short-form "what do you do" → "asking millionaires how to make a million dollars" + yellow text / outstretched hands / shocked homeowner. 5M views; the business-YouTube door-knock template.

Keep a brain pile of outliers the way a 10% body-fat room changes dinner. Pull a 6-year-old viral when you need a move.

6. 100 → 10 → 1 (Jobs / Apple)

Paddy: boutique team, 2–3–500 ideas/month per client; client sees ~5. 100 ideas to get 10; 10 to pick 1.

Elimination criteria (4–5 gates, must be strict):

Feasible (cannot put a car on the moon; Elon may not be bookable).
Can it be titled and thumbnailed? Kills ~2/3. "How do we put this in a thumbnail?" → dead.
Aspirational view target (Open Residency ~100k combined → reject a 20–30k guest if the goal is 150–200k).
CCN / 80% overlap.
ICP / guest fit (their Cloud skill: distribution × communication × ICP).

If 100 ideas → 30 survive, criteria are garbage. If 100 → 1–2, too strict.

For the last 10: one-pager each — rough thumbnail sketch, titles, log line, skeleton. Nine never shoot. Team (or solo, gun-to-the-head): which one would you bet your career on?

Start with packaging first, then the interview. A nine-figure friend who cannot be packaged is not worth the slot.

7. Packaging workflow (title → brief → shoot → design → publish)

Titles first (not thumbnail). Minimum 10 titles. World's fastest camera drone vs F1 / featuring Verstappen / Max races drone — five variants in five seconds.
Thumbnail brief before the shoot. Five concepts, angles, pose. Hand sketch or Gemini mock. Do not shoot the video then ask "what's the thumbnail?"
Shoot the stills. Hundreds of frames, micro-expressions, lighting, 10–30 shots per concept. (Phone: 11 poses in a few seconds.)
Design ≥3 fully different thumbnails from that source. 45–50 is what they do on monster videos (five designers). Three is the contract floor. Shirt-color swaps are not three thumbnails.
Native YouTube A/B on the three thumbs. Pick the one title you would bet a career on; leave it unless the video misses expectation.

Click-to-unpause: thumbnail looks like the video already started (cars on the grid, mid-speech with complementary text). A posed smile + thumbs-up is not unpause. Do not double the same POV in title text and thumb text.

Red Bull "world's fastest camera drone vs F1" (Verstappen): superlative + F1 = fastest-on-earth association + pause-at-the-start-line + leading lines + clean + famous face not required in the thumb. Competing package on the same idea was estimated 1M vs 10M+. Now 30M+. Adapted from CarWow lineup films; others later race Usain Bolt the same way.

Nonlinear returns: 20–30% better thumb can be 40× views (Gabe web-designer: 1k/day → 40k/day on an already live video). NBA: 10% better in the right ways ≠ 10% more money.

After you have a recognizable brand: incremental thumb changes (keep guest POV + color; add objects) so you do not dump equity. Open Residency Peter Rahal / David Protein lawsuit objects on the left.

Last-video halo: Outdoor Boys can screenshot and still get clicks. That is earned, not a beginner strategy.

8. After publish (expectation, not vibes)

Before upload, the team predicts hour / day / week views and a 1-of-10 rank vs last 10.

Elon should be a 1/10. If it opens 2/10, panic — change thumb / stop the test and go dramatic. Squid Game opening 3/10 is a miss vs expectation even if "above average."
A video you already knew was small that opens 8/10 is fine — do not "fix" it.
Good data > no data > bad data. If it is beating expectation, do not shake the title. If it is missing, no-data on an untried title is better than sitting on bad data. Do not retitle 10 times in day one — you cannot attribute.

Sometimes you still post a lower-view tweet/video (program promo) for upside. Make that rare. "View maximizer" is the default; niche accountants who need 1,000 yacht buyers are the exception, and they still play 1-of-10 against themselves.

9. Title principles

Fewest words that stay interesting. Soft cap ~60 characters (TV shows more than mobile). Under-50 is a relic, not a god.
Universal language. "D1 athlete" and "Goodwill" shrink TAM outside the US. "Pro athlete." EBITDA in a broad title is 5/100 people on the street. Diary of a CEO uses profit / money. Jargon is OK only if you want that qualified slice.
Complement the thumbnail, do not repeat it. "How anyone can go viral in 2026" + thumb "this is the key" = curiosity. Same sentence twice = dead.
Readable at a glance. Short contemporary words ("modern" not "contemporary"). No three-titles-glued-together. Title Case (not ALL CAPS — equal letter height, worse recall).
Do not SEO-optimize the soul out of it. Most channels: search ~5–10%, subs box <5%, channel page ~3%. "2026 iPhone 16 review unboxing tutorial" loses to "Don't waste your money reviewing the new iPhone 16" (same keyword, human interest).
Good → better → best: "Best value Porsche GT3 on the market" → "I found the best value…" (story) → "I found the cheapest 911 GT3…" (superlative) → "I found the cheapest 911 GT3 on the market, but nobody will buy it" (contradiction). Inverse: worst apartment in NYC that's expensive but everyone wants in.

Frameworks get you 0→80. 80→100 is curiosity, taste, and a real idea.

10. Retention and the 45-second intro

Intro ≤ 45s on most winners. Simple language, intrigue, proof. Three steps:

Deliver the promise (title/thumb is a promise). Visual stronger than verbal. Drone-vs-F1 that does not show a drone or car in 10s: he has seen 60–70% gone in 10 seconds. "You made the right click."
Create intrigue with least context. Not a table of contents. Open hooks you will pay off. Drip-feed names and lore (MrBeast: "FBI agent, 24 hours to catch me, go" — not a biography). "At the end of the day" is ending language — people leave.
Seamless flow. Best intro does not feel like an intro. Connecting sentence into the first beat. Highlight reel can work at this production level (Diary of a CEO); Modern Wisdom often cold-opens on a question. Pick a lane; shorten as you grow.

Stakes make structure matter. Paper-in-the-bin: no one cares → camera op says you can't → $1M + punch → FOMO. Types: enemy, time limit, reward, punishment, cost of not paying attention. Match the video: slime-on-the-host does not belong on "bulletproof investment portfolio"; "$3M retirement gap if you pick the average one" does.

Retention curves drop as a video reaches colder audiences; shape (where they leave) stays. Study thousands of curves for "this phrase = 2% leave."

YouTube is evergreen. Feed mixes 8-year-old and 2-hour-old. The algorithm mostly follows the audience (including TV: 15–20% of client traffic → 50–60% in five years because people want long videos on a TV, not because YouTube "prioritized" them). Shorts on the phone, long on the TV. Small creators still win if the video is better; AI is dissolving editor/graphics gates; small = more room to experiment without a core to disappoint.

11. Brands, pendulum, AI, decade

Do not start a channel you will not run. Half-ass brand YouTube (OpenAI dumping ads) is worse than no channel. Hire a 22-year-old who already grew something ($80–100k), not a pretty-graphics person with zero YouTube. $100k budget → that hire, or Paddy 10 years ago.

Pendulum:

First: watchable (audio/picture not broken). He has seen a $400B public brand that failed this.
Then ideas + packaging.
After packaging is good, swing back to production (4K, extra angles). Brands usually obsess "make it look better" and skip title/thumb/story.

Distribution after the video: UTMs and evergreen ad reads still sell years later (example: "Kao" in Austin). Cars: ~90% of US car purchase research now happens on YouTube. CAC vs LTV of a Honda-for-life buyer. Commit 18–24 months.

AI: use in pre (Claude + vidIQ/1of10 MCP for outlier brainstorms; Gemini for thumb concepts). Writing/scripts still human on the last 20%. Do not skip guest-prep.

YouTube next 10 years (his bet): TV-length premium vs raw 30–40 min rants vs Shorts (and kids watching both at once). Beating Netflix/Disney on watch time because the model is creator-enabled, not curated. Skill of virality leaks into film (Backrooms, Obsession), politics (Mamdani, Trump podcasts), sport (Lamine Yamal YT vs a La Liga scorer nobody has heard of). Ads + sponsors = phase 1 for 95% of creators; digital products phase 2; CPG (Feastables, Jolly Rancher-class) phase 3. Everything is still downstream of views.

Study ritual: playlists named outlier / title-thumb / retention / intro. Add while watching. 3–5 scheduled hours/week to rewatch and note. The hard skill is again, in multiple niches, not one viral.

12. Operator checklist (ship this, not a vibe)

Before you green-light a video:

[ ] Log line passes the glance test (one second).
[ ] CCN: core, casual, and new can watch without missing homework.
[ ] 80% overlap with the channel through-line (or this is the 20% swing).
[ ] 100-idea funnel ran; elimination criteria killed the unpackable ones.
[ ] ≥10 titles, ≥3 differentiated thumbs, title chosen, native A/B ready.
[ ] Thumbnail stills briefed before cameras roll.
[ ] Intro ≤45s: promise, intrigue, flow. Stakes match the format.
[ ] Expected 1-of-10 written down before publish.
[ ] After live: views + curve. Change package only vs expectation, not boredom.

Free companion from the episode: openresidency.com/paddy-galloway-playbook.