SCOREConsumer & creative
Jev: Viral Post Analysis System
Kiyoro@0xKiyoro
0:15Jev is a system that analyzes viral X posts by classifying each post into 14 yes-or-no categories, extracting features like hook type and confidence score, to identify effective content elements for a writing model named Astra.
TakeHook type, first-line number, proof quality, open loop and confidence score are extracted in one parallel pass.
Jev analyzed 100,000 viral X posts for $0.67.
Astra only had to write from what survived.
That separation is the entire system.
Instead of making a writing model read the complete corpus, Jev turns every post into structured data through 14 yes-or-no classifications. Hook type, first-line number, proof quality, open loop and confidence score are extracted in one parallel pass.
The run finished in 20.4 seconds.
Claude Opus 5 received the same corpus and clock. It processed 214 posts for $0.98. At the observed rate, analyzing all 100,000 would have cost roughly $458.
Jev also found that the most common hook was the weakest:
> Superlative claim: 2.34% viral
> Contrarian take: 1.59%
> Launch or tool drop: 1.46%
> Numbered list: 0.55%
Astra receives those findings and writes the final post around the strongest combination of hook, proof and structure.
The writing model no longer searches through 100,000 examples.
It starts with the answer.
Jev finds what works. Astra turns it into something worth publishing.
