Jev for instant compaction
Post on X arguing that agent context compaction should be a Jev keep-or-drop decision rather than a summarisation prompt.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
This X post by Tamara Tran argues that agent context compaction should be handled as a Jev decision, deciding what to keep, rather than by prompting an LLM to summarise. The author also maintains fast-jev-compaction, which jev-compaction-plus forked.
Only the list summary was collected, so the full argument is not reproduced here.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Should this piece of context be kept?(inferred) | — |
How it works
- Treat each context item as a keep or drop decision
- Ask Jev instead of generating a summary
- Keep chosen items verbatim
Strengths
- Avoids generated summaries that may invent details
Caveats
- Summary only; post content not collected
At a glance
- Type
- Discussion
- Status
- discussion
- Has code
- no
- Has tests
- no
Use cases and integrations
How curated lists describe it
X: argues agent context compaction should be a Jev decision rather than a summarization prompt.
yibie/awesome-jev
Frequently asked questions
What does the post argue?
That compaction should be a Jev decision, not a summarisation prompt.
Is there code?
Related implementations exist, such as fast-jev-compaction, but this entry is the post.
Where is it?
On X.
Related projects
X post by Theo arguing that decision-model context compaction is reconstruction, not filtering, and is a poor strategy.
DiscussiondiscussionAuto-updated bilingual leaderboard of Jev ecosystem projects, tutorials and discussions, discovered daily and sorted by GitHub stars.
ProjectproductionChatGPT-style web chat where every word of every reply is a Jev multiple-choice decision over a ~1,800-word vocabulary plus END.
ProjectprototypeTypeScriptExperiment that deliberately misuses Jev as a text generator by asking which symbol comes next and sampling from the returned distribution.
ProjectresearchPythonChat client that makes Jev write by asking Next word? over a 20,000-word dictionary in 255-word buckets plus a run-off.
ProjectresearchPythonChinese translation of 'Jev Engineering for coding agents', design notes on a harness where Jev makes every per-turn decision.
Projectdiscussion