This is the most hardcore real-world use case I've seen so

@NFT_Chen SuSu_酥酥👅 · Sep 18, 2026 · case #091174 · listed via Jev Hub

Still from the clip posted with this case; watch it in the original post on X.

Used Jev to score each tool call in a 156k-token Claude Code session and drop stale reads and noise, compacting it to 62k tokens without rewriting.

The original post, translated

💥Holy crap! This is the most hardcore real-world use case I've seen so far for the Jev model! Instant Compaction only deletes, doesn't rewrite. In the demo, a 156k tokens Claude Code session, Jev scores each tool call one by one, instantly compressing it to 62k. The user's goal, exact error messages, and key code text are all still there; expired Read / intermediate noise directly DROP. The skills Jev truly has here: 🔹For each

Show the original post in its source language ▾

💥我靠！这是目前看到的给 Jev 模型找到的最狠的真实用例！ Instant Compaction只删，不改写。演示里 156k tokens 的 Claude Code 会话，Jev 逐条给 tool call 打分，瞬间压到 62k。用户目标、精确报错、关键代码原文全在；过期 Read / 中间噪音直接 DROP。 Jev 在这里真正会的技能： 🔹给每条

The post above is a machine translation from zh; the untranslated text is in the fold-out.

Engagement when collected

Views68K

Likes12

Bookmarks0

Reposts0

Replies0

Quotes0

Numbers are a snapshot taken from X when the case was added to the library (schema v1, collected 2026-09-19); they will not match today.

Where this case fits

Filed under agents & workflows, coding & developer tools. In the pattern Jev is built for, the model answers a bounded question per step — and ordinary code acts on the answer, because the answer is already a value rather than a paragraph. Other posts in the same family are on the agents & workflows page.

🤖 Agents & workflows 🛠️ Coding & developer tools