用 Jev 打分 + Grok 4.7 解读爆款推文
作者用 Jev 对 64,800 条爆款推文各问 12 个结构化问题打分，31 秒完成 77.7 万次决策，再让 Grok 4.7 只解读前 1%，总结出爆款规律。
i just built an X viral post viewer with Grok 4.7 + Jev and it read more posts in 31 seconds than i did all year jev does the sorting. grok 4.7 does the explaining. the whole run cost less than a bus ticket x link → jev asks 12 questions → score → grok 4.7 opens only the top 1% → why it went viral jev doesn't write anything. it only answers simple questions about each post: number in the hook, yes or no. which format, pick one. how baity, score 1 to 10. it can't make things up, because it can only pick from answers i allowed i gave it 64,800 viral posts from my niche. 12 questions each is 777,600 decisions, done in 31.4 seconds for $0.93 reading them by hand at 20 seconds a post is 360 hours of scrolling. people would say 1,000x faster. i did the math and it's 41,274x that's $0.000014 per post, and grok 4.7 only reads the 648 worth reading what came back: 71% of the winners have a number in the first line. 83% are under 200 words the big mistake: pasting 50 viral posts into the smart model and asking "why did this work". it reads 50 and guesses. jev reads 64,800 and counts setup took me 11 minutes: x api key in the secure field, one schema with 12 questions, one grok skill, test on 100 posts first jev scores, grok 4.7 explains, and i still write the post myself everyone is asking the smart model how to go viral. the cheap one that reads everything already knows
这条是引用着另一条写的：x.com/razeden0/status/2101981166500815075