⚡ GitHub Pulse · 早报

生成于 2026-09-19 03:46 UTC · 追踪 2,154 仓库 · 8 多源共振
分化明显:coder/coder、hypit 多源共振且有真实开发活动(substance);而登顶的 'jev' 集群全是 hype-ish,警惕刷榜。
🎯 查清 'jev' 集群为何集体登顶却零开发活动(疑似刷榜),同时把 coder/coder 加入重点追踪名单。

⭐ 多源共振

今日两极:coder/coder 与 hypit 被多源确认且 substance(有 PR/issue),是真动量;jev-ultrafast、security-audit-skill 虽 4 源但 hype-ish——光涨星、几乎没有开发活动。

githubxsocialboard
+335★/d hype-ish official #8
A coding-agent skill for multi-phase security audits with independently verified, machine-readable findings
🔺 @MaciejLukianski 首发 · 41h 前
githubxsocialboard
+317★/d hype-ish official #7
🔺 @betterhn20 首发 · 16h 前
官方 #1 但 hype-ish(几乎零实质开发)+ 'jev' 集群同时暴涨,强烈警惕刷榜。
coder/coder ×4 🆕 new
githubxsocialmodel
+309★/d substance
🔺 @LFrefman 首发 · 13h 前
多源共振 + substance + 关联模型;面向开发者 agent 的安全环境,真需求,今日第一优先。
deeplethe/utopia ×4 🆕 new
githubxsocialboard
+290★/d substance official #2
Local-first, agent-assisted document-to-ontology workbench
sapientinc/PRAXIST ×4 🆕 new
githubxsocialboard
+205★/d substance official #4
Autonomous research system for measurable, computer-executable research.
trycua/cua ×4 🆕 new
githubxsocialboard
+189★/d substance official #3
Scale computer-use 2.0 with open-source drivers, cross-OS fleets, and benchmarks for training, evaluation, and data generation.
🔺 @trycua 首发 · 6h 前
latent-spaces/brag ×4 🆕 new
githubxsocialboard
+140★/d hype-ish official #15
You built it. Now brag. Turn the project you just created into a short, shareable launch video with one command.
hypit-ai/hypit ×4 🆕 new
githubxsociallaunch
+130★/d substance
🔺 @cccyd_qwq 首发 · 31h 前
多源 + 有真实开发活动;病毒视频克隆,同款产品也在 whatships 发布过——工具+话题双热。

🔥 动量榜

榜首被 'jev' 集群(jev-ultrafast / fast-jev-compaction)占据,全部 hype-ish 且同时暴涨,疑似协同刷榜;真正值得看的是 alibaba/open-code-review(+1714,substance)。

#Repo7d+1d★7d★质地官方X
1 cloudflare/security-audit-skill 🆕 new
JavaScript · A coding-agent skill for multi-phase security audits with in
🔺 @MaciejLukianski 首发 · 41h 前
+335 10,636 hype-ish #8 27×
2 browser-use/jev-ultrafast 🆕 new
Python
🔺 @betterhn20 首发 · 16h 前
官方 #1 但 hype-ish(几乎零实质开发)+ 'jev' 集群同时暴涨,强烈警惕刷榜。
+317 5,600 hype-ish #7
3 deeplethe/utopia 🆕 new
Python · Local-first, agent-assisted document-to-ontology workbench
+290 1,685 substance #2 10×
4 robbietilton/Compositor 🆕 new
Swift · The Photoshop alternative for Mac
🔺 @dotey 首发 · 5h 前
+260 460 hype-ish #1
5 sapientinc/PRAXIST 🆕 new
Python · Autonomous research system for measurable, computer-executab
+205 970 substance #4 10×
6 tamaratran/fast-jev-compaction 🆕 new
TypeScript · Claude Code plugin that replaces the compaction summary with
+196 2,365 hype-ish #11
7 trycua/cua 🆕 new
HTML · Scale computer-use 2.0 with open-source drivers, cross-OS fl
🔺 @trycua 首发 · 6h 前
+189 967 substance #3 10×
8 arcboxlabs/arcbox 🆕 new
Rust · Run AI agents on real and isolated machines — own kernel, fi
+168 803 hype-ish #6
9 eternity4719/HowToLiveBetter 🆕 new
HTML
🔺 @knowledgefxg 首发 · 60h 前
+147 4,962 hype-ish 10×
10 latent-spaces/brag 🆕 new
Python · You built it. Now brag. Turn the project you just created in
+140 3,753 hype-ish #15 10×
11 alibaba/open-code-review 🆕 new
Go
🔺 @shao__meng 首发 · 71h 前
真动量:+1714 且开发活动扎实(substance),阿里出品的代码审查工具。
+133 14,046 substance 22×
12 hypit-ai/hypit 🆕 new
TypeScript
🔺 @cccyd_qwq 首发 · 31h 前
多源 + 有真实开发活动;病毒视频克隆,同款产品也在 whatships 发布过——工具+话题双热。
+130 9,940 substance 25×
13 deepseek-ai/deepseek-harness 🆕 new
TypeScript
🔺 @the_osps 首发 · 7h 前
+120 7,640 hype-ish 10×
14 dramaclaw/dramaclaw 🆕 new
TypeScript · A general-purpose AIGC video engine: script to finished film
+116 465 substance #5
15 MiniMax-AI/minimax-code 🆕 new
TypeScript · An open-source coding agent for your terminal, powered by Mi
+108 679 hype-ish #14
16 ayghri/i-have-adhd 🆕 new
Python
🔺 @zhtyyx 首发 · 69h 前
+107 4,712 hype-ish 42×
17 stablyai/orca 🆕 new
TypeScript · Orca is the ADE for working with a fleet of parallel agents.
🔺 @GitTrend0x 首发 · 62h 前
+105 4,728 substance #22 20×
18 affaan-m/ECC 🆕 new
JavaScript
🔺 @BlockInsight214 首发 · 10h 前
+101 5,039 substance 14×
19 tt-a1i/archify 🆕 new
JavaScript
🔺 @GitTrend0x 首发 · 62h 前
+99 7,261 substance 12×
20 Tencent/BrowserSkill 🆕 new
Rust · Let AI agents use your real, logged-in browser without inter
🔺 @JackAIStudio999 首发 · 44h 前
+99 3,407 substance #23 16×

🧠 模型发布时间线

优先看 source_count≥2(多源确认)与 ★notable 的;单源的谨慎。

日期模型来源
2026-09-17KAT-Coder-Pro V2.5
Kwaipilot
aimlapi
2026-09-17Gemini Omni Flash Preview
Google
aimlapi
2026-09-17Venice Uncensored
Venice
aimlapi
2026-09-17Nano Banana 2 Lite
Google
aimlapi
2026-09-17Jev 1.13
TypeSafe AI
aimlapi
2026-09-16Union Alpha
Stealth
aimlapi
2026-09-12Schematron V2 Turbo
Inference.net
aimlapi
2026-09-12Schematron V2 Small
Inference.net
aimlapi
2026-09-11Fugu Ultra v2.0
Sakana AI
llmgateway
2026-09-11Kimi K2.8 Preview
Moonshot AI
llmstats
2026-09-11Atria Dawn Preview
Shanghai AI Laboratory
llmstatsllmgateway
2026-09-11Fugu Ultra v2
Sakana AI
aimlapi
2026-09-11Fugu Max
Sakana AI
aimlapillmgateway
2026-09-10Ling 3.0 Flash VL ★
inclusionAI
aimlapiopper
2026-09-10DeepSeek V4.1 Flash ★
DeepSeek AI
aimlapillmstatsopperllmgateway
2026-09-10DeepSeek Chat (V4.1 Flash)
DeepSeek AI
aimlapi
2026-09-08GPT Image 2.5 Sunburst
Open AI
aimlapillmgateway
2026-09-08GPT Image 2.5 Flare
Open AI
aimlapillmgateway
2026-09-08Mercury 2.5
Inception
aimlapi

🤗 HF 采用榜

Modeldownloadslikespipeline

🗞️ Hacker News

📄 论文

Recursive Self-Improvement / Coding Agents 主题论文近期集中出现,值得留意方向。

Haocheng Xi, Yiming Xie, Hexu Zhao · 2026-09-19
Video diffusion models repeatedly process long spatiotemporal token sequences during denoising, making attention a major computational bottleneck. Linear attention offers an appealing alternative and has been widely adopted in recent large
Video generationarXivpwcarxiv
Dongzhou Cheng, Taoran Yi, Ye Fang · 2026-09-19
Enabling robots to adapt to unfamiliar environments as readily as humans remains a moonshot goal of embodied AI. No finite collection of demonstrations can cover every task and situation a robot will encounter, making the ability to learn f
RoboticsarXivpwc
YiFan Zhang, Yunheng Zou, Shaokun Zhang · 2026-09-19
Autonomous research loops such as AutoResearch show that one coding agent can improve a training setup unattended. Run several of them and each session starts from scratch, so more agents tend to mean more duplicated search rather than more
Coding AgentsarXivpwc
Hejia Geng, Zesen Huang, Haoyang LI · 2026-09-19
Scientific code repositories encode decades of human knowledge in executable models, methods, and tools. Yet fragmented toolchains, implicit domain conventions, and specialized correctness criteria make this knowledge difficult to convert i
AgentsCoding AgentsReinforcement LearningarXivpwc
Meng Luo, Yanlin Li, Hao Li · 2026-09-19
Foundation models, alongside advances in learned game-world models, are reshaping AI across the game lifecycle. Beyond playing games, recent systems model players and game dynamics, support design and development, adapt player-facing experi
World ModelsarXivpwc
Honglin Guo, Tao Gui, Yicheng Chen · 2026-09-19
As AI agents become participants in the development of their successors, they reshape both the production of intelligence and the role of human researchers. We introduce Atria Dawn Preview, a foundation agentic language model designed for s
AgentsCoding AgentsComputer Use AgentsDeep Research AgentsarXivpwc
Haiwen Diao, Jiahao Wang, Chenjing Ding · 2026-09-19
We launch SenseNova-U1.5, an 8B-MoT native unified multimodal model that understands, reasons about, and generates visual content within an encoder-free and VAE-free architecture. We strengthen its visual interface through spatially coheren
Image editingImage generationImage UnderstandingInstruction FollowingarXivpwc
junyan ye, Wei Liu, DONGZHI JIANG · 2026-09-19
While diffusion base models such as GPT-Image-2 and Nano-Banana exhibit remarkable visual expressiveness, their end-to-end generation inherently yields flattened bitmaps with error-prone text, precluding layer-wise post-editing. Conversely,
Coding AgentsImage editingImage generationarXivpwc
Xionghao Wu, Yijun Yang, Shiyang Zhou · 2026-09-19
Robotic manipulation faces a fundamental scaling challenge: robust generalization demands broad physical experience, yet action-labeled robot trajectories are expensive to collect and inherently limited in diversity. Egocentric videos offer
RoboticsWorld ModelsarXivpwc
Hanyang Wang, Yimo Cai, Weiliang Chen · 2026-09-19
Physical understanding and reasoning depend on forming compact and generalizable representations of the world. While modern vision-language models can recognize and explain diverse physical events, they often lack explicit representations o
ReasoningWorld ModelsarXivpwc

📢 电报精选

🎯 Alpha 账号

作者leadslead率仓库数
@shanyanggm140.6421
@shaw_stone7383270.76
@xzbx88870.788
@GitTrend0x70.78
@FrontieraTechIT50.717
@DataChaz514
@the_osps50.835
@xfubot50.57
@LoveAIbrain514
@iasg100450.3610
↑ 顶部