Stop the thumb

LiveFresh finds

This part of the guide writes itself. A scheduled job reads what changed across Shorts, Reels, TikTok, X and LinkedIn, then files each thing below with the number behind it, whether anyone saw it coming, and why it happened.

1First, the honest part

You asked me to study 200 viral Shorts. I can't watch videos, so pretending I did would be the first bad hook in this guide. Instead, I went through studies where other people measured thousands of short videos, added what the platforms themselves have said, and mixed in your own teardown of 15 long-form hits. Together that covers far more than 200 videos. Every number in this guide says where it came from, and I'll tell you when a source is weak.

Trust dots: three dots means the platform said it or the data is big and careful. Two dots means a big dataset from a company that sells a tool. One dot means a rough benchmark.

On top of those, I used the rules the platforms have shared themselves: YouTube's help pages and product updates, what Instagram's head Adam Mosseri has confirmed, and the feed code X published in January 2026.

Three brutal truths before we start

So why go all in?

Because Shorts are the cheapest hook lab on the planet. You can test 30 openers a month for almost no money, and every test teaches you something about your audience. Your buyers are also a small, specific crowd, so for fromSilicon, 500 of the right founders and investors watching beats a million random people.

Your Shorts have three jobs

2How the Shorts machine decides

When you post a Short, YouTube doesn't show it to everyone. Picture a restaurant that serves your new dish to one small table first. If that table eats it, the kitchen sends it out to a bigger table, and then a bigger one. If the first table pushes the plate away, the dish goes back to the kitchen and stays there.

Every table is a test, and the test watches a few simple numbers. Learn these four and you can read any Short's report card.

What broke? A quick doctor's check

3The hook: the Receipt Hook

After all that data, one hook shape keeps showing up near the top, and it matches the rule in your own long-form rubric. You show the proof first, say something people would bet against, and crack it open before they can swipe. I call it the Receipt Hook, because a receipt is proof that something really happened.

Why each part works

• Show. In the 4,000-video study, "proof drop" openers (a chart, a receipt or a specific number) got the most saves by far, about 1,761 per video. Videos where the creator stood in front of the source material, the greenscreen format, averaged about 150,000 views, compared with about 56,000 for a plain talking head.

• Say. Bold "hot take" and "investigator" openers averaged about 140,000 views. Openers that start a story slowly, like "So the other day...", averaged about 7,000. That's roughly 20 times worse.

• Crack. Your rubric says the viewer should never reach a resting place. Your long-form hits change direction with a "but" about every 26 seconds, and a Short has to land its first crack in about 3.

Four Receipt Hooks, taken apart

4What 4,000 videos say

This is the biggest recent study I found: 3,997 TikTok and Instagram Reels videos from 109 creators, tagged by an AI and compared by average views. It isn't YouTube, and the company that ran it sells a tool, so treat it as a map of where to test and not as a law. Still, some of the gaps are huge.

Pick the emotion by the job

5How long should a Short be?

This is where the studies disagree, and the disagreement teaches you something useful: length follows the idea and the platform.

The rule: use the shortest length that lets the payoff land. Extra seconds are something you earn by holding people, and filler never earns them. A joke told too long dies, and a story told too short makes no sense, so let your retention graph tell you which mistake you're making.

6After the hook: the body

A great hook with a boring middle is a trailer for a bad movie. The body has one job: keep paying off the hook, one small surprise at a time. Here is the beat map for a 35-second Short.

Change something every 3 to 5 seconds: a cut, a zoom, new text or a new fact. Your long-form hits change direction about every 26 seconds, and a Short needs that same heartbeat running about five times faster.

The loop trick

Write your last line so it runs straight into your first line. When the Short restarts, the sentence finishes itself, and many people watch the start again before they notice.

7The hook bank: 12 patterns

Top creators rotate a small set of hook shapes and swap the topic. Here are 12 that fit your three lanes: Receipts (startup and VC stories), Clips (moments from client podcasts) and Build (autoBlade in public). Spin the machine when you're stuck, or scroll the full list under it.

8The kill list

These openers die in the first second. Most of them are habits from TV and old-school YouTube, and the thumb has zero patience for any of them.

9Swipe or stay?

Ten openers. Decide fast, like the thumb does, and then see why.

10The all-in Shorts system

Going all in means running three lanes on a fixed weekly rhythm, with a scoreboard. Posting random Shorts every day until one hits is how people burn out in a month.

The biggest jump comes from going from 1 or 2 posts a week up to 3 to 5, and follower growth more than doubled at that level. More posts still helped, just a bit less with each step. On LinkedIn, 2 to 5 posts a week was the sweet spot, worth about 1,182 extra impressions per post compared with posting once a week. X is the odd one out, because a per-creator cap means posting more there spreads your reach thinner.

The weekly machine

1. Research on Monday morning. Pick 7 topics and find the receipt for each one before you write a single word.

2. Hooks first. Write three first lines for every topic, score them with the scorecard below, and keep the best one.

3. Batch film in one two-hour session. Seven Shorts, one setup, one shirt.

4. Edit fast. Run clips and build videos through autoBlade, and do Receipts as greenscreen over the source.

5. Post native. Upload the real file to each platform, with captions burned in and no watermarks.

6. Score at 48 hours and again at 7 days. Log every hook in your sheet.

7. Recycle. Winning patterns get new topics, and losing Shorts get a new first line and a second chance on another platform.

One file, five platforms, five sets of rules

The scoreboard: your hook log

This sheet turns your taste into a system, and it becomes the most valuable thing you own after 30 Shorts. Sort it by viewed vs swiped away every Sunday and look at which patterns sit on top.

1110 hooks ready to film

Seven Receipts and three Build hooks. The facts are real, so you can film them this week, but check every number against a primary source before it goes on screen. A wrong receipt is worse than no receipt.

1. Receipts

   "Instagram had 13 employees when Facebook paid $1 billion for it."

   Receipt: the April 2012 deal headline with the team size. Crack: "So what was Facebook really buying?"

   Copy hook
2. Receipts

   "WhatsApp had 55 employees when Facebook paid $19 billion for it."

   Receipt: the 2014 deal headline. Crack: "That works out to about $345 million per employee."

   Copy hook
3. Receipts

   "Airbnb stayed alive in 2008 by selling boxes of cereal."

   Receipt: a drawn card of the cereal boxes and the year. Crack: "And that cereal is part of why Y Combinator let them in."

   Copy hook
4. Receipts

   "Juicero raised about $120 million for a juicer."

   Receipt: the funding number next to the machine. Crack: "Then reporters squeezed the juice packs by hand and got nearly the same juice."

   Copy hook
5. Receipts

   "Quibi raised $1.75 billion and shut down about six months after launch."

   Receipt: the launch month and the shutdown month side by side. Crack: "So where does $1.75 billion go in six months?"

   Copy hook
6. Receipts

   "Blockbuster could have bought Netflix for $50 million in 2000, and it said no."

   Receipt: the year and the price on screen. Crack: "And at the time, saying no looked like the smart call."

   Copy hook
7. Receipts

   "OpenAI's board fired Sam Altman on a Friday, and within five days they agreed to bring him back."

   Receipt: a Friday-to-Wednesday calendar strip. Crack: "More than 700 employees had threatened to quit."

   Copy hook
8. Build

   "It took me three nights to edit one podcast episode, so I'm building an app that does it in about twenty minutes."

   Receipt: a timer running over the autoBlade timeline. Crack: "The hardest part turned out to be figuring out who's talking."

   Copy hook
9. Build

   "Three microphones in one room all hear the same person, so figuring out who's talking is the hardest thing my app does."

   Receipt: three audio waveforms spiking together. Crack: "Guess wrong and it cuts to someone sitting in silence."

   Copy hook
10. Build

    "My app learned to find laughter by accident."

    Receipt: two audio bars spiking at the same moment. Crack: "When two mics spike at once, somebody just laughed."

    Copy hook

For the Clips lane, run your hook finder script on each client transcript and pick lines a stranger would argue with. Those hooks have to come from real episodes, so there are none to invent here.

12Score your hook before you film

This is your long-form rubric, squeezed for Shorts. The weights are my starting guess, so log each score next to the real numbers, and after about 30 Shorts, keep what predicts results and drop what doesn't.

Bands match your long-form rubric: 80 and up, ship it. 65 to 79, fix the weakest part. 50 to 64, rewrite the first line. Under 50, pick a new idea.

13The bigger game: long-form YouTube

Shorts get you seen. Long videos build trust, pay better and turn strangers into people who know your name. Your documentary series is where fromSilicon's reputation gets made, and the long-form machine plays by different rules.

• YouTube ranks long videos on how satisfied people feel afterward, not only on minutes watched.

• Test & Compare lets you test up to three titles or thumbnails on one video, and it picks the winner by watch time. Run it on every upload.

• YouTube put cutting AI slop on its 2026 priority list, so channels that remix the same public facts are getting squeezed. Your edge is access: real investors and founders on camera, which your client work already gives you.

• Crowded topics need a receipt nobody else has. Sam Altman's story has been told by a lot of channels, so your version needs an interview, a document or a person who was in the room.

• Use Shorts as your title lab. When a Short's first line wins, try it as a long-form title in Test & Compare.

14X: the dinner party

X works like a dinner party. Nobody listens to the stranger shouting from the doorway, and everyone listens to the guest who has been talking with them all night.

• X published its Grok-powered feed code on GitHub in January 2026, so a lot of the guesswork is gone.

• A bookmark counts about ten times as much as a like, and a reply that you answer counts far more than both.

• There's a cap on how many of your posts one person sees, so posting ten times a day just spreads you thinner.

• Since a July 2026 update, your posts reach people you follow each other with much more often than strangers.

• Posts with links get about 30 to 50 percent less reach in most analyses. One reading of the code disputes that, but putting the link in the first reply costs you nothing.

• The ranker leans heavily toward Premium accounts, so pay for it.

15LinkedIn: the conference hallway

LinkedIn is the hallway at a conference. People are there to learn from someone in their field, and that's exactly where emerging fund managers and founders hang out between meetings.

• It rewards clear expertise in a niche, and it buries generic AI-written posts.

• Your connections and followers see your posts first, so who you're connected to matters more than any trick.

• Buffer's study of over 2 million posts found 2 to 5 posts a week is the sweet spot, worth about 1,182 more impressions per post than posting once a week.

• Upload your Shorts directly as native video instead of linking out to YouTube, because every platform prefers to keep people on its own site.

16Making autoBlade go viral

For a product, the demo is the ad. Nobody shares a feature list, and people do share a result they can't quite believe.

The build rhythm

17How to upsell yourself

A chef with no reviews has to charge less, no matter how good the food is. Your channel and your numbers are your reviews.

"Distribution strategist" is a bold title, and your own rubric has a rule for bold claims: no proof object, no license. So go collect the proof.

• One flagship hit on your own channel. Investors will check it before they read your pitch.

• Analytics access on every client channel, so you can show before-and-after numbers.

• Your hook log with 30 or more tests. Almost nobody in VC content has one, and it turns your taste into a system people can pay for.

• Case studies with real numbers, in a shape like "We changed the first line, and viewed vs swiped went from 52 to 74 percent." That's an example shape, so use your real numbers.

The ladder

18The 30-day all-in sprint

Tick things off as you go. Your progress saves in this browser only, and the thumb gets more impressed as you go.

19Words you'll see

Hook

The first one to three seconds of a video. Its only job is to stop the thumb.

Receipt

Proof on screen, like a number, a headline, a screenshot or a timer.

Crack

The "but", stake or question that makes the viewer need the next few seconds.

Viewed vs swiped away

The share of people who stayed instead of swiping to the next Short.

Engaged views

Views from people who stayed past the first few seconds. Since 2025 this is closest to the old meaning of a view.

Average percentage viewed

How much of the video people watched on average. Loops can push it past 100 percent.

Loop

An ending that runs straight back into the opening, so people watch again.

RPM

How much you earn for every 1,000 views after the platform takes its cut.

Greenscreen

You on camera in front of the source, like an article, a chart or a clip.

Trial Reels

Instagram Reels shown to non-followers first, which makes them a free hook test.

Test & Compare

YouTube's tool for testing up to three titles or thumbnails on one video.

Mutuals

People on X who follow you while you follow them back.

Retainer

A monthly fee for ongoing work, instead of a price per job.

CTA

Call to action, like "follow for more". The data says to let the payoff do the asking.

20Sources

Vendor studies are marked. Platform rules change often, so check anything that matters before you bet money on it.

1. The Content Labs: 3,997 TikTok and Reels videos (vendor study)
2. Antoine Lacombled: a study of 3.3 billion Shorts views
3. Shortimize: Shorts view counting and engaged views
4. Prepublish: reading viewed vs swiped away
5. Retensis: Shorts retention benchmarks (vendor benchmark)
6. Piktochart: Inflow Network's 5,400-Short length study
7. Argil: Vidooly's 2 million+ Shorts length study
8. Adobe Express: what the most-viewed Shorts share
9. vidIQ: Shorts pay and RPM
10. AIR Media-Tech: Shorts RPM across 274 channels
11. OutlierKit: YouTube algorithm updates, including the Shorts split
12. Miraflow: 2026 Shorts changes, search and dislike tests
13. SocialBee: YouTube and viewer satisfaction
14. Search Engine Journal: Test & Compare for titles
15. Creator Weekly: YouTube's 2026 priorities
16. SocialPilot: the X algorithm in 2026
17. FORKOFF: X signal weights
18. AdLibrary: links and reach on X
19. Postory: the case against the link penalty
20. Sprout Social: Premium and the X feed
21. SocialPilot: Instagram Reels signals and Trial Reels
22. Dataslayer: Instagram in 2026
23. Buffer: how often to post on Instagram
24. Buffer: how often to post on LinkedIn
25. Metricool: the LinkedIn algorithm
26. Buffer: who sees your LinkedIn posts first
27. Your video-packaging skill: the 15-video teardown and scorecard, plus your build-in-public script rules for the autoBlade facts.