用 Grok+Jev 搭自动选题新闻室，日省 3 小时
作者用 Grok 抓 X 爆款、Jev 做 skip/keep/draft 决策门，人工只做终审，声称一条 Claude 帖跑到 210 万浏览。
this X setup is starting to feel f*cking unfair. I basically built a tiny newsroom where Grok hunts viral X posts, JEV decides what deserves attention and I only show up at the end to approve. before this my “system” was: open X -> scroll forever -> save 20 posts -> check numbers -> dump ideas into notes -> somehow still have no idea what to post. ~3 hours gone. so I replaced myself with a little AI desk: -> Scout watches ~18 accounts. replies get ignored. under 5K views gets buried. 10K+ gets starred -> then it leaves my bubble and checks ~20 fresh AI posts from the last 48 hours to catch topics suddenly popping up everywhere -> JEV becomes the gate. instead of making a big model think about every random post, it can make the cheap little calls first: skip / keep / investigate / draft -> the survivors get dissected: hook, topic, keywords, structure, tool, CTA -> Draft turns those patterns into a fresh English post + an idea for the video or visual -> Board remembers everything: what’s hot, what’s drafted, what I posted, what still needs work then I come in. > read it. > kill the dumb stuff. > rewrite a few lines. > hit Post. nothing auto-posts. and that’s probably my favorite part: the system isn’t trying to magically predict what will go viral. it watches what people are already voting for with their attention - then works backward. one Claude bug-hunting post from this workflow reached around 2.1M views. roughly what this gives me now: ~3 hours/day back ~$1K/month I don’t spend on assistants 1–2 posts/day without spending my life scrolling X > Grok scouts. > JEV filters. > the writer writes. > I make the final call. basically, I accidentally built myself a tiny AI media team. and I made the whole thing public + free.
这条是引用着另一条写的：x.com/thegreatest_sv/status/2100236496837984387