data-backed patterns

Hook Formulas

Reusable patterns behind top-performing posts — each one derived from, and backed by, real hooks with named creators, sources, and documented performance. No formula exists here without real posts behind it.

Real hooks. Real creators. Real receipts.

you never have to guess on a hook again. every hook, formula, and breakdown is pulled from real posts that already performed, across X, YouTube, and TikTok, each with a named creator and a source. metrics only show when a source documented them. nothing is made up.

Twitter/Xup to 7M views · 24.3k likes · 5.7k RT

Specific result + timeframe

Concrete numbers make a claim believable and the 'how' irresistible — a dollar figure, a metric, a timeframe, an age.

The game community in shambles, throwing tantrums and melting down because an indie hacker built a ${amount}/mo game in a week is worth of studying.

21 real posts · 19 with metrics
Twitter/Xup to 51.9M views · 160.6k likes · 13.2k RT

I spent N days/hours doing X

Commit to an extreme amount of time or repetition on one thing. The concrete number makes the feat feel real and the outcome impossible to ignore.

It’s only been {number} days since OpenAI revealed GPT-4o.

9 real posts · 8 with metrics
Twitter/Xup to 14.8M views · 192.4k likes · 10.3k RT

Earned authority (I analyzed / I've done N)

Lead with a number, credential, or body of work, then trade on it for an insight the reader couldn't get elsewhere.

Google Docs is used by {number}+ billion people worldwide.

9 real posts · 9 with metrics
Twitter/Xup to 872.1k views · 8k likes · 726 RT

Question hook

Mirror the viewer's own curiosity with a question they can't confidently answer, opening a gap only the content closes.

How exactly do Language Models perceive time?

9 real posts · 9 with metrics
Twitter/Xup to 3.4M views · 16.8k likes · 3.9k RT

Contrarian / false belief

Challenge something the audience quietly believes, or state a blunt reframe. The dissonance stops the scroll and invites debate.

If you're not using AI, you're falling behind in {number}.

6 real posts · 5 with metrics
Twitter/Xup to 7.7M views · 103.7k likes · 16.5k RT

Before → After transformation

Anchor a painful or humble 'before' against a desirable 'after'. The visible gap promises a payoff worth watching to the end.

How I went from {relatable low point} to {aspirational outcome}.

5 real posts · 5 with metrics
Twitter/Xup to 5M views · 23.5k likes · 5.1k RT

Curiosity gap / forbidden knowledge

Hint at valuable information without revealing it — a secret, a hidden reason, something 'they don't want you to know'. The open loop pulls the read.

Noise is a secret destroyer of productivity.

5 real posts · 5 with metrics
Twitter/Xup to 319.3k likes · 60.4k RT

Numbered list (N things / mistakes)

A countable, finite promise of dense value. A specific number sets expectations and signals it won't ramble.

{number} free {resources} that are so valuable they feel illegal to know:

5 real posts · 5 with metrics
Twitter/Xup to 9M views · 44.4k likes · 7.3k RT

Command / pattern interrupt

A direct command interrupts autopilot scrolling and implies the viewer is currently doing something wrong.

Education will never be the same.

5 real posts · 5 with metrics
Twitter/Xup to 8.9M views · 27.2k likes · 1.9k RT

Simple stakes + big reward

A dead-simple premise anyone understands, attached to a large or absurd payoff. The stakes do the persuading.

{Famous person} often talks about {action} using this one {thing}.

4 real posts · 4 with metrics
Twitter/Xup to 4M views · 52.9k likes · 6.7k RT

X vs Y (extreme comparison)

Put two starkly different things side by side. The contrast creates an instant curiosity gap about how they actually compare.

I made the closed-source vs. open-weight models figure for this moment.

4 real posts · 4 with metrics
Twitter/Xup to 956.2k views · 14.5k likes · 3.3k RT

Mistake / warning

Loss aversion: warn the viewer they might be doing something wrong. Fear of the error keeps them watching to check.

your data is safe. we don't train on it. because it would make the model worse

4 real posts · 3 with metrics
Twitter/Xup to 109.1k views · 973 likes · 32 RT

Callout to the exact audience

Name the specific person you're talking to so the right viewer feels personally addressed and stops scrolling.

Every once in a while, it hits you that word2vec and attention were one year apart.

2 real posts · 2 with metrics