Captions, hooks and scripts

Now the daily craft: turning the calendar's one-line ideas into finished posts. This is where volume actually lives — captions, hooks, and Reel scripts, week after week — and where AI assistance goes from convenient to structural, because the work is formulaic in exactly the way models excel at and the formula genuinely matters. Feed content lives or dies in its first second, and that second has a name.

The hook is the whole game

A caption's first line and a Reel's first two seconds do one job: interrupt the scroll. Platforms truncate captions after the first line or two; viewers swipe Reels in under a second — everything after the hook exists only if the hook works. So generate accordingly: never ask for "a caption"; ask for hooks first, in bulk.

The post: our master printer Ramniwas-ji carving a new wooden block —
it takes him 3 days, the design outlives him by decades. Give me 10
hook lines, mixed: some curiosity, some bold statement, some question,
some sensory detail. No clickbait — it has to be true to the video.

Ten arrive in seconds. Perhaps two are good, five are workable, three are filler — and that's the correct outcome: you're using the model as a variety engine (it can explore angles far faster than you) while you stay the judge of true-to-the-brand (lesson 1's division, in miniature). Pick the hook, then ask for the rest — "build the caption on hook 3: two short paragraphs, the 3-days detail, end with a soft question, voice doc applies" — and the structure follows the classic arc: hook, story, close. For Reels, the same conversation produces a shot-by-shot script: hook line on screen in the first two seconds, beats of the story with on-screen text per shot, a close that invites a comment. Meera's Friday batch session (lesson 3) is exactly this loop run five times — under an hour for the week.

Format is a dialect

One idea does not fit all placements — each format has its own grammar, and the model speaks all of them fluently if asked: the feed post gets the full story; the Reel script needs spoken-language rhythm and on-screen text cues; the story sequence is three tappable frames ending in a poll; the carousel opens with its strongest claim and reveals one point per slide. Never port text between formats by pasting — re-ask: "same idea, as a 7-slide carousel" produces structure, not a chopped-up caption. (The full one-idea-to-ten-assets chain gets its own lesson shortly.)

Two guardrails, learned the hard way

Verify every fact against your own knowledge, every time. The model drafting Meera's knowledge-pillar post about block-printing history wrote fluent, confident, plausible paragraphs — with one invented date and a technique detail no printer would recognise. Models produce plausibility, not truth (the companion course explains exactly why), and in your domain of expertise you are the fact-checker — that's not a burden, it's your credibility doing its job. An expert audience unfollows on the first fake detail.

Keep the last-mile edit. Before anything ships, one human pass: cut the line that's trying too hard, add the word only you would use, check it sounds like you said it. Thirty seconds per post — the difference between "content" and "your content", and the moat from lesson 1, maintained daily.

Do this today: take this week's next planned post and run the full chain — ten hooks, pick one, build the caption, then the same idea as a Reel script. Time yourself: the whole chain runs under fifteen minutes, and that number is why this course exists.