WHAT NEW IN X11.05 THU

X 上的 AI 合集档案 · 23 个合集 · 69,408 条

jevVIDEO

Awesome Jev:896 个仓库的 Jev 生态地图

有人把 Jev 及同类模型的公开项目整理成清单:60 个本地模型、220 个 agent 工具、73 个浏览器/电脑操作项目,展示「LLM 出选项、Jev 选一个、代码执行」的分层用法。

Rexei
@iamrexei2026.09.25收录原推发布于 2026.09.25
Rexei @iamrexei原推发布于 2026.09.25
在 X 上打开原推

HOW THE F**K IS THERE ANOTHER JEV TOOL? Yet another Jev tool But this isn't just another local clone, an API wrapper, or a 30-line MCP implementation "Awesome Jev" has compiled a map of the entire ecosystem surrounding Jev and Jev-like models At the time of writing, it lists 896 public repositories: > 60 local and open Jev-like models > 220 agent tools > 73 projects for browser and computer use > 55 SDKs and clients > 113 benchmarks and research studies In other words, Jev launched quite recently, yet the following are already emerging around it: ▸ local models like Laya ▸ servers based on DiffusionGemma and vLLM ▸ routers that decide which tool to assign to an agent ▸ validation, moderation, and risk assessment layers ▸ browser agents that require a quick solution rather than text This no longer looks like just a "cool closed-source model that outputs JSON" A distinct layer of AI infrastructure is emerging: > LLM proposes options → Jev-like model selects one → code executes the action And that, it seems, is the key takeaway. We’re used to assigning generation, reasoning, and decision-making all to the LLM. But now, we’re seeing many attempts to offload that last part—decision-making—to a separate, fast, and cheap layer. But let's keep it real: 896 repositories don't equate to 896 solid products The list includes early experiments, forks, SDKs with barely any stars, and bold claims lacking independent benchmarks The catalog itself is independent and unaffiliated with TypeSafe Still, as a map showing just how fast the open-source ecosystem around Jev is growing, it’s one of the most useful finds I’ve come across lately

这条是引用着另一条写的:x.com/iamrexei/status/2102783484636930263

x.com/iamrexei/status/2103197460676182236采集于 2026.09.25 10:08