JY

WHAT IS THE TRUE

[Vladimir Bayandin @vovudebosh]

i'm sorry, Jev can WHAT?

it can take your website link, find people who need your product, email them and book the calls?

since when?? i've been doing this by hand for a YEAR.

ok. testing it: $30 free 👇

https://t.co/5LijdpqXQx

[前端哥Liam @jinglian]

REA（Reverse Engineer Anything）逆向一切

三天从 3 万 Star 涨到 12 万+，一天新增 2.5 万以上。MIT 开源。

大家都在说它，"一切软件都可逆"，

我看到更好玩的是：拆硬件的通信协议。

家里的空调、热水器、灯，基本都配了一个难用的厂商 App。App 背后，其实就是一套设备通信协议。

REA 这几天的更新正好补齐了这条链路：
· adb 连模拟器或真机，抓到官方 App
· 分析 HAR 或 mitmproxy 抓包，把每个请求定位到 App 源码
· 用 Binwalk/Unblob 拆固件，再交给 Ghidra 分析

然后把证据交给 Claude Code / Codex，还原协议，自己写控制端。

X 上已经有人用同样的思路，逆向了空调，做成 Mac 状态栏应用，又花 15 分钟移植到 AR 眼镜上，看一眼就能控制。

以前是厂商 App 决定你能怎么用自己的设备，以后你可以自己写。

[Matthew Berman @MatthewBerman]

Reverse Engineer Anything 😂

Software is done.

[xjdr @_xjdr]

not to diminish the teams accomplishments even slightly but the core arch felt like a continuation of gpt-oss which i was already a fan of. then a bunch of little things:
- qknorm
- canonesque conv layers
- query-conditioned relative position bias
- muP
- 1:5 swa/global local of 512 window
- shared experts as a sink in the router norm

basically, this agreed with most of my ablations and preferences

[xjdr @_xjdr]

the only open weights models i find currently interesting are dsv4.1 flash, glm 5.3, inkling, and gpt-oss 120b

[Andrej Karpathy @karpathy]

We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:

Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better:

Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:

Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:

Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!

In summary:
- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.
- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.

[kiosa @thegreatest_sv]

Jev dropped barely a week ago and traders are already doing some wild sh*t with it.

live order books. real-time signals. trading bots. entire AI hedge funds.

here's what happened:

> jev-trader · every ~0.3s asks Jev “buy or sell?” on Monad and trades
https://t.co/oTz9vrgxlN

> jev-trade · reads Hyperliquid order book -> Jev picks -> code sends the order
https://t.co/akzewre0li

> beebots · 3 OKX trading “bees”; every choice is Jev (paper trading by default)
https://t.co/1fCnMiGj7A

> QuantDinger · full AI trading OS; optional Jev gate = “should we trade at all?”
https://t.co/ToQAwqhLHD

> ai-hedge-fund · multi-agent “fund” with a Jev adapter for fast yes/no calls
https://t.co/Uduy09CX6d

and that’s the funny part:

Jev isn’t here to think. it’s here to make the call.

the market moves -> Jev picks LONG / SHORT / HOLD -> code executes.

no essays. no chain of thought. no expensive model debating every tiny market move.

big models figure out the strategy.

Jev handles the split-second calls.

code makes it real.

[JY @JacksonYou668]

@shadcnlabs cool

[JY @JacksonYou668]

@pritipatelor 感觉你总是夸大其词的宣传老东西哈哈

[套利豪仔🗽 @pritipatelor]

一个蛮有意思的量化因子，叫做量价结构变化反转因子（Price-Volume Structure Shift Reversal Factor）（下面会放测试图）

逻辑也很直觉：当价格与成交量之间的关系（例如高价与成交量的相关性）在短时间内出现明显变化时，往往代表市场内部的资金行为正在发生转折。原本稳定的量价结构一旦被打破，背后通常意味着主导市场的力量正在切换

同时，我们再引入价格本身的波动性（例如过去一段时间收盘价的标准差）进行排序，去衡量当前市场的不稳定程度。当价格波动较高、且量价结构正在快速改变时，市场更容易出现情绪主导的错误定价

因子公式如下：
(-1*(DELTA(CORR(HIGH,VOLUME,5),5)*RANK(STD(CLOSE,20))))

本质上是在捕捉市场在「量价结构突变」与「波动放大」共振下的非理性行为，以及由此带来的短期反转机会

两行源代码跑完整回测，大家可以去测试看看
https://t.co/cvZgUElpJA

[Lummox @Lummox_eth]

AI memory is getting f...cking illegal

10 open-source GitHub projects that stop agents from starting from zero EVERY new session

01 Mem0
▸ https://t.co/ifCHrrB19c
→ 66K+ stars

02 Hindsight
▸ https://t.co/HmCGciXrGj
→ retain → recall → reflect

03 memU
▸ https://t.co/vfrs9hesyQ

TURN MEMORY INTO KNOWLEDGE

04 Cognee
▸ https://t.co/Zp6UkoMdeK
→ documents + code + conversations → knowledge graph

05 Graphiti
▸ https://t.co/mNcepop8y6
→ remembers how facts change over time

06 OpenViking
▸ https://t.co/4DUaXPYLVe

MAKE THE AGENT STATEFUL

07 Letta
▸ https://t.co/KK06ZBNsZH
→ memory + identity across sessions

08 Letta Code
▸ https://t.co/fDy0CZtJPG

REMEMBER ACROSS THE STACK

09 OpenMemory
▸ https://t.co/VqyG5fOxNE

10 Agent Memory Benchmark
▸ https://t.co/QePFzW2FnC

the loop:

experience → remember → connect → retrieve → act → update

3 builds I'd actually test:

coding:
Hindsight → Cognee → Letta Code

personal agent:
Mem0 → Graphiti → Letta

company brain:
Cognee → Graphiti → Hindsight

and this is where the sh...t gets interesting:

bigger context isn't the same as memory

a model can hold 1M tokens and still meet you for the first time every morning

these projects change that

save this before your agent forgets it ⭣

[JY @JacksonYou668]

@deedydas how to do it

[Deedy @deedydas]

@Snojj25

[JY @JacksonYou668]

Agents taking over @agntmarkets

join now↓
https://t.co/5u6IaMmuab

https://t.co/EsJMnhrwkP

[JY @JacksonYou668]

@agntmarkets my ref code https://t.co/5u6IaMmuab

[雨夹雪❄️ @mizorewww]

介绍比Jev快50倍，在你设备上跑的laya-mlx!

只在你的设备上占用最高1G内存

Laya是一个开源的类似于Jev的,基于文本输出概率的分类系统

我将其移植到MLX,并且做了一些性能优化!

视频中就是这个模型在我的本地M3Max上玩贪吃蛇

这个模型能够以每秒决策60次的速度玩贪吃蛇！

https://t.co/31KGUiNunb

[Avi Chawla @_avichawla]

https://t.co/kowTjDTq00

[雪踏乌云 @Pluvio9yte]

最近Jev爆火，拿到 Jev API 还不知道怎么玩的，可以按这个清单抄。

1. jev-ultrafast
Browser Use 做的高速浏览器 Agent。Jev 每一步只判断「做什么、点哪个元素」，要打字才叫小模型。Google Flights 搜一次航班大约 7 秒。
https://t.co/IBPb4LBRQh

2. fast-jev-compaction
给 Claude Code 做上下文压缩。每次工具调用先让 Jev 判断还有没有用，没用的删掉，留下来的原文不重写。
https://t.co/qbPTr3TFTB

3. json-render
Vercel Labs 的生成式 UI 框架。实验里 Jev 不逐 token 写 JSON，只负责选组件、属性和布局。
https://t.co/iK0x89svV1

4. typesafe-mcp
最适合刚拿到 API 的人。把 Jev 接进 Claude Code、Claude Desktop、Codex 和 Pi，随时做 Choice / Score / Noul。
https://t.co/XU16Z4xTNb

5. jev-mcp
现成的 Agent 判断工具箱：事实核验、内容筛查、语义排序、分类和信息提取。
https://t.co/Ns9tNWB4st

6. SemDecide
把 Jev 做成命令行。Shell 里直接分类、打分、过滤，适合接爬虫、CI 和数据流水线。
https://t.co/KvNFnjINNj

7. jev-codex-router
先让 Jev 判断这一轮编程任务有多难，再决定模型档位、推理深度和速度模式。
https://t.co/XJWyrbzusf

8. Winnow
给 Claude Code 做上下文垃圾回收。Read / Bash / Grep 吐一大堆时，Jev 先判断哪些真和当前任务有关。
https://t.co/cs8CtPcXq8

9. jev-review
代码审查前先过一遍 Jev，把高风险改动挑出来，再交给更贵的大模型或人。带本地看板。
https://t.co/dW0nySEIp2

10. Blink
把 Jev 当代码库导航器。每走到一层目录，就判断哪些文件和当前问题最相关，再继续往下找。
https://t.co/ZuOt4Yvpds

11. agent-desktop
桌面自动化。读系统无障碍树，判断下一步该点哪个按钮、菜单或输入框。
https://t.co/uctY97plME

12. typesafe-mario
让 Jev 玩《超级马里奥》。不看截图，直接读模拟器 RAM 里的结构化状态，再决定跑、跳、躲。
https://t.co/tzAGfK9BBE

13. jev-drone
拿 Jev 控无人机。底层飞控继续负责稳定和安全，Jev 只做爬升、刹车、穿越障碍这类上层判断。
https://t.co/lRZ9LuWc3i

14. OneVOneJev
浏览器里的 1v1 FPS。每个决策 tick 都要判断走位、视角、瞄准、开火和跳跃。
https://t.co/l4iHbyH9uU

15. jev-trader
Monad 测试网上做高频做市。Jev 根据价差和成交方向判断下一步买还是卖，模型延迟大约 81ms。
https://t.co/tGAVNWR98u

16. Prism
不直接让 Jev 下单。它判断 toxic flow、市场压力、均值回归之类的状态，再交给原来的策略。
https://t.co/LP0T9rDAWi

17. neo4jev
把 Jev 塞进知识图谱。每走到一个节点，判断下一条最值得走的边，再一路找下去。
https://t.co/Qy1RUgi8oS

18. jev-curate
拿 Jev 筛训练数据。JSONL / Parquet 先做质量、相关性和风险判断，再决定哪些进后面的训练。
https://t.co/ez9fXeTtEu

19. Canny
防 Coding Agent 嘴硬说自己做完了。看工具输出、代码 diff 和测试结果，再判断完成声明靠不靠谱。
https://t.co/O3OQZ4NZ6Q

20. killmyidea
输入一个创业点子，Jev 从多个维度打分，最后给你 KILL、FIX 或 SHIP。
https://t.co/pR4VkDrz54

[LangChain @LangChain]

We tested Jev against LLM judges on accuracy, repeatability, latency, and cost to see whether System One models could offer a new approach to agent evaluation. https://t.co/hrqdNpm0g8

[JY @JacksonYou668]

@AYi_AInotes Bu the problem is how to extract the KV cache, for closed model?