👁 160 次阅读
HackerNews 今日热榜
| rank | title | score |
|---|---|---|
| 1 | OpenLogi | 623 |
| 2 | Being ambitious and being a dad | 483 |
| 3 | A 3D fruit fly on macOS desktop powered by the real FlyWire connectome | 272 |
| 4 | Cerebras CS-4 | 267 |
| 5 | Scientists stunned by children’s lung recovery in ultra low emission zone | 247 |
| 6 | Palomar: A registry of Lean verified mathematics | 98 |
| 7 | [Supersonic Trebuchet ](https://www.youtube.com/watch?v=Co57SfcT-h0 | 97 |
| 8 | Where Human Sleep Went Wrong | 46 |
| 9 | λλ: A Programming Language for Silicon Photonics | 18 |
| 10 | The Two Factions of C++ | 14 |
GitHub 热门AI项目
| rank | repo | stars | description |
|---|---|---|---|
| 1 | seyedehsanhadi/sloptrim | 162 | A local detector for AI-writing patterns. Scores every prose |
| 2 | vuphongle/oss-pr-reviewer | 111 | AI-powered CLI for reviewing GitHub pull requests, detecting |
| 3 | thiientv/godmode | 89 | Production-grade Agent Skills for AI coding agents—composabl |
| 4 | wtfzambo/speak-like-you-eat | 51 | Translate AI garbage to human language. A Pi package for pla |
| 5 | Greninja9257/LabLLM | 50 | A native macOS lab for teaching tiny language models to thin |
| 6 | diyanbogdanov/emend | 43 | Verified AI dependency migrations. Finds which of your lines |
| 7 | Copilot">diegosouzapw/OmniCopilot | 41 | 1200+ AI models in your GitHub Copilot Chat — free & forever |
| 8 | elie222/botdirectory.ai | 40 | Open-source directory of agent-bot prompts for Grok Bot, Rak |
| 9 | nexpeakcore/deepseek-harness-pr-review | 40 | AI code review with DeepSeek: headless PR review automation |
| 10 | PerryLink/dsh-auto-review | 40 | Second-model AI auto-review for DeepSeek Harness approval re |
Arxiv 最新AI论文
| rank | title | authors |
|---|---|---|
| 1 | On the Fragility of Self-Improving Agents: Variance, Task Or | Qinyuan Ye, Yu Li, Yada Pruksachat |
| 2 | Against Political Polarization: A Unified Framework for Trac | Yijie Xu, Chao Wang, Hui Xiong |
| 3 | Efficient RLVR Scheduling via Graph-Structured Online Diffic | Zhizhao Liu, Zhiliang Tian, Xi Wang |
| 4 | Grading Needs a Rubric, Not Intelligence | Jhen-Ke Lin |
| 5 | BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Be | Liubov Chubarov, Alexandra Kules, Daniil Volkov |
| 6 | Encoded but Not Actionable: Auditing the Decode-Generate-Ste | Man Liang, Xinzhao Cheng, Faizan Wajid |
| 7 | Interpretable Humans, Alien LLMs: Expert Analysis of Latent | Alona Strugatsk, Licol Zeinfeld, Jason Cooper |
| 8 | What Aggregate Scores Miss: Measuring Item-Level Regressions | Xiaonan Xu, Wenjing Wu |
| 9 | LLM-Derived Preference Judgments Are Not Self-Consistent | Matthew T. Ford, Francis Bahk, Jingjing Wang |
| 10 | MoNe: Modular Neural Memory for Efficient Long Context Infer | Wonguk Cho, Kyubyung Chae, Tribhuvanesh Or |
自动采集于 2026-08-19
🔗 相关阅读（你可能还想看）
📄 Gemini Omni 1.2 Flash 发布：视频降价30%、原生15秒，单条成本只贵5分钱
谷歌 11 月 5 日开放 Gemini Omni 1.2 Flash：720p 视频每秒 0.07 美元，比上代降 3
📄 GPT-6 Cyber 发布：121 万美元，948 条通往 root 的路
OpenAI 发布 GPT-6 Cyber：在 Linux 内核上跑出 948 个本地提权利用，整轮 API 成本 12
📄 Anthropic冲刺IPO：170亿营收、8950亿算力承诺、首次在招股书披露递归自我改进风险
📄 Anthropic IPO招股书曝光：单季营收170亿美元背后的8954亿美元算力账单
📄 阿里千问接入 Apple Intelligence：中国版 iPhone 的 AI 时刻来了
📄 苹果AI中国落地靴子落地：千问接入Apple Intelligence，11月4日开测
📄 苹果官宣：Apple Intelligence 国行版落地，阿里千问深度接入 Siri 与系统写作工具
苹果官宣 Apple Intelligence 国行版 11 月 4 日开测、12 月随 iOS 27.3 正式上线。阿
📄 Anthropic 递交 S-1 启动 IPO：Claude 母公司估值剑指 2 万亿美元
📄 Cerebras 公开万亿参数模型部署方案：59 块晶圆跑 3T 参数，推理 2000 tokens/秒