CampusAI-app/campusai-video-kit

Hook & retention psychology

The single highest-leverage thing in short-form is the first two seconds. Reach is cheap; attention is the whole game. A great idea with a weak opening dies at 1.5 seconds and the algorithm never shows it again. This skill makes the opening earn the next two seconds, and the body keep them, honestly.

The one rule

Earn attention, don't bait it. Every tactic below has an honest version and a cheap version. The cheap version (fake claims, rage bait, "wait for it" with no payoff, misleading thumbnails) buys a first view and burns trust. Retention craters, and it poisons the brand. We do the honest version every time: the curiosity is real, the payoff lands, the promise in the hook is kept in the body. This is the honest-product rule applied to attention.

How attention actually works (the mechanism)

• The feed is a rhythm. People scroll in a trance, half-deciding every ~0.8s "keep or flick." You are not competing with other content for a decision; you are trying to break a reflex.
• Curiosity is a loop that wants closing. The brain hates an open question. An opening that raises a specific question buys the seconds it takes to answer it. Vague questions don't (the brain shrugs); specific ones do.
• Prediction is death. The instant a viewer can predict the rest, they leave. Retention is a series of small "I didn't see that coming" beats that reset the prediction.
• Relevance is a filter, not a hook. "This is for me" stops the right people; it doesn't create interest on its own. You still need a reason to watch.

The hook: five levers (use 1–2, not all)

Write ten openings, keep three, post one. Each opening should pull at least one lever hard.

1. Open loop: raise a specific question the video answers. "The one setting every Smith freshman changes in week one" (not "here are some tips"). The loop must be closed later, or retention dies on the next post too.
2. Stakes / cost of ignoring: what they lose by scrolling past. "Most people miss the career-fair slot that actually matters." Loss reads louder than gain.
3. Specificity: a real number, name, or detail makes it true and un-scrollable. "I checked 40 events this week; 3 are worth your time" beats "check out these events." Specific = credible = watchable.
4. Pattern interrupt: break the feed's rhythm in the first frame: an unexpected visual, a mid-motion start, using the app's own UI as the gag (see the burger-shop scroll-stopper). The thumb pauses to figure out what it's looking at. Native beats polished.
5. Direct callout: name the exact person. "If you're a junior who still hasn't touched your resume..." The right viewer feels addressed; everyone else is fine to lose.

First-frame rules: front-load the payoff-question, not a logo or a slow intro. No "hey guys, welcome back." The first spoken words and the on-screen text should both carry the hook (many watch muted). If the first 3 words are throat-clearing, cut them.

Retention: keep them past the hook

• Open fast, mid-motion. Start on the second sentence. Cut the runway.
• Change something every ~2–3 seconds: shot, framing, caption, zoom, b-roll. Visual change resets attention before it drifts.
• No dead air. Every beat either advances the point or raises a new mini-question. Cut breaths, "ums," and setup that doesn't pay.
• Nest loops. Open a small new question before you fully close the last one, so there's always a reason to stay.
• Deliver the promise, then a little more. The hook wrote a check; cash it clearly, then over-deliver by one beat so the ending earns a save/share.
• Payoff timing. If the whole thing is a single "wait for it," the wait must be short and the reveal must be worth it. Long teases with thin payoffs are the fastest way to train the algorithm to bury you.
• End on a reason to act or re-watch, not a fade. Saves and shares are the strongest signal; give them a line worth saving.

Diagnose a struggling post

• Reach but no watch-through → the hook is the problem. Rewrite the first 2 seconds; test 3 new openings.
• Drop at ~3s → the first frame raised no question, or the promise felt generic. Add specificity or stakes.
• Drop in the middle → it got predictable or slow. Add a loop or a visual change where the graph dips.
• Watched but no saves/shares → the payoff was fine but forgettable. Sharpen the ending into one quotable/savable line.
• Good on one platform, dead on another → native mismatch. Re-cut the opening to that feed's rhythm.

Scoring rubric (0–2 each; a post ships at strong on hook)

• Hook: specific question or stake in the first 2s? (0 none / 1 vague / 2 sharp and specific)
• First frame: stops the scroll on its own, muted? (visual + text both carry it)
• Honesty: is the curiosity real and the promise kept? (any bait = automatic fail, rewrite)
• Retention: change/beat every 2–3s, no dead air?
• Payoff: promise cashed, plus one, savable ending?
• Relevance: right viewer feels addressed without narrowing to nothing?

Any bait/misleading element fails the whole post regardless of other scores.

How this plugs into our system

• campusai-video is the method for making the film; this skill is the depth behind its hook and pacing. Run it on the opening before building, and again on the near-final cut.
• Human reference to study: Jonathan Maxim (content & retention psychology). Craft reference for scroll-stopping first frames: the burger-shop feed-glitch Reel.
• Never invent a stat, credential, or claim to strengthen a hook. If the specific detail isn't true, find one that is. There's always a real, specific angle.