The 'AI Built My Nephew's Game' Hook and Four Other Formats Winning This Week

YouTube17 HN points

'Fable 5.2 one-shotted my nephew's dream game' — personal story format where a creator uses a family member or friend as the protagonist to show what a new AI model actually did, step by step. The structure is: setup (who asked, what they wanted), the prompt, the output, the reaction. No benchmark slides, just a screen recording and a voice-over.

Why it works: A named non-technical person as the test subject makes the capability feel real and relatable rather than abstract. The emotional payoff (nephew's reaction) gives the video a natural ending.

YouTube819 HN points

Paper-airplane tutorial revival — foldableflight.com hit 819 HN points with zero marketing copy, just clean step-by-step fold diagrams. Creators are filming 90-second hands-on tutorials of physical skills (origami, knots, paper crafts) with overhead phone mounts. The hook is always a finished object in frame before the first fold.

Why it works: Physical how-to content has near-zero drop-off because the viewer needs to pause and follow along. Overhead angle removes the need for editing cuts.

Instagram209 HN points

Satellite imagery carousel — watched.earth shows every satellite pass over any location this week with actual photos. Creators are screenshotting their city or a famous landmark, pairing it with a ground-level photo, and running a 5-slide carousel: slide 1 mystery crop, slide 2 reveal, slides 3–5 context facts.

Why it works: The mystery-crop opener forces a swipe. Local angle makes the content feel personally relevant to followers in that city.

LinkedIn324 HN points

AI privacy concern post — the Google Gemini / human annotator story (324 HN points, 76 comments) is generating strong engagement on LinkedIn as creators post a 3-paragraph take: what the report says, what it means for their workflow, one concrete thing they changed. No outrage, just a calm decision log.

Why it works: Decision-log format positions the creator as thoughtful rather than reactive. It invites comments from people who have made the same or opposite call.

X724 HN points

Model behaviour change thread — the 'Is Opus 5.5 nerfed?' discussion (724 points, 468 comments) shows creators and developers posting before/after prompt comparisons as image threads. Format: tweet 1 is the claim, tweets 2–4 are screenshots of the same prompt run on different dates, tweet 5 is the open question to the audience.

Why it works: Before/after evidence is inherently shareable. The open question at the end converts readers into repliers, which boosts algorithmic reach.