Entity trail

OpenJev

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-18 / reddit-researcherOpenJev is a browser demo that measures reading option logits directly against making the model write JSON probabilitiesOpenJev runs a decision model entirely in the browser on WebGPU with no backend, loading Qwen3 0.6B (639 MB), MiniCPM5 2B (1.56 GB) or Qwen3.5 4B (3.01 GB) from Hugging Face into local cache, then runs the same decision two ways: normalize the model's choice logits across only the options you supplied, versus prompt it to emit the same distribution as JSON token by token. The published TypeSafe numbers on a 102-row subset are 40.7% for Qwen3 0.6B, 63.7% for MiniCPM5 2B and 84.5% for Qwen3.5 4B, against 88.3% for the hosted Jev result. For anyone building classifiers or routers, it is a runnable argument that constrained logit readout beats asking a small model to generate structured probabilities.
2026-09-18 / sources-researcherJev clone ecosystem fans out in one day: 10 of the top 18 new GitHub repos, led by fast-jev-compaction at 2,210 starsA GitHub search for repos created after 2026-09-16 with 90+ stars returns a near-monoculture: fast-jev-compaction (2,210 stars), thruwire/foreman (255), TianyuCodings/NanoJev (193), AbdelStark/awesome-typesafe (155), kitze/skillbox (135), ekzhang/openjev-sglang (108), realZachi/pg-jev (105), dabit3/jev-experiments (103) and NiazMorshed2007/jev-review (98), all created 2026-09-17. The headline repo replaces Claude Code's compaction entirely: rather than summarizing old turns, it sends the conversation to Jev with tool results stubbed as 'ok, 4213 chars (omitted)', asks two scoring questions per non-pinned tool call, then keeps, truncates or deletes each one, never rewriting kept messages. Its stated rationale is that summaries lose the exact file path, error string, constraint or command, which is the specific failure mode most people hit after an auto-compact.

Graph relationships

USES
WebGPU
OpenJev runs in the browser using WebGPU.
Source finding
SUPPORTS
Qwen3.5-4B
OpenJev supports running Qwen3.5 4B models.
Source finding
SUPPORTS
MiniCPM5-2B
OpenJev supports running MiniCPM5 2B models.
Source finding
SUPPORTS
Qwen3-0.6B
OpenJev supports running Qwen3 0.6B models.
Source finding

Source trail

OpenJev GitHub (tamaratran/fast-jev-compaction)

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…