2026-10-14 AI 日报
统计日时区：America/Los_Angeles · 已结算
morluto/rea · vxcontrol/pentagi · ifixai-ai/iFixAi
开源项目
- morluto/rea
Reverse engineer anything with agents, from app behavior down to native binaries.
- vxcontrol/pentagi
Fully autonomous AI Agents system capable of performing complex penetration testing tasks
- ifixai-ai/iFixAi
Independent Auditing of AI Agents. Run by human or the agent itself, to answer the most crucial question in the AI Agent Economy. Is the agent doing what is…
- K-Dense-AI/scientific-agent-skills
Turn any AI agent into an AI Scientist. The #1 Agent Skills library for science, used by 250,000+ scientists worldwide. 177 ready-to-use validated skills plus…
- mksglu/context-mode
Context window optimization for AI coding agents. Sandboxes tool output (98% reduction), persists session memory, and enforces routing across 17 platforms via…
- heygen-com/hyperframes
Write HTML. Render video. Built for agents.
- stablyai/orca
Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop, mobile and remote runtime.
- tashfeenahmed/freellmapi
7.4 billion tokens per month. 34 free LLM providers. 635 free model endpoints. All behind one /v1 endpoint, plus any custom OpenAI-compatible endpoint. Smart…
- anthropics/claude-plugins-official
Official, Anthropic-managed directory of high quality Claude Code Plugins.
- xerj-org/xerj
XERJ is the new way for AI to search data. Its autoindex capability activates agents to know your data without the token waste of grep and sed. One command…
论文
- Jeva: System One Models from First Principles
System one models enable direct, low-latency decision-making. Yet recent practice of predicting an answer from serialized text does not, by itself, respect…
- VideoJev: System One Model for Video Understanding and Interaction
Visual agents must turn video observations into timely decisions to track progress and choose what to do next. Building on the emerging System One paradigm,…
- Urbex: Agentic Spatial Grounding for City-Scale 3D Scenes
City-scale 3D grounding is commonly formulated as candidate ranking, which requires precomputed instance proposals and becomes inefficient in large outdoor…
- LeWAM: End-to-End World and Action Modeling with JEPAs
Joint-embedding predictive architectures (JEPAs) learn compact latent world models without pixel reconstruction. Existing JEPA pipelines typically train…
- Harness^2: Recursive Agent Harnessing for an Open World
LLM agents' capabilities depend not only on their underlying models but also on their execution harnesses. Recursive self-improvement (RSI) offers a promising…
- Vela: User Behavior as Supervision Sparks LLM Intelligence
Large language models are increasingly deployed in applications that act on users' long-term behavior histories, from personal assistants to recommendation…
- Multics: Generative Multi-Player World Model with Hierarchical Communication
Multi-player interactive world models aim to simulate shared environments from multiple players' perspectives, but achieving robust generation while scaling…
- Agent-Reflex: Teaching Language Agents to Act Reflectively
Large language models have evolved from single-turn chatbots into interactive agents capable of long-horizon reasoning. In real deployments, MCP tool services…
- TPPO: Tracking-Value PPO For Language Model Reasoning
Proximal Policy Optimization (PPO) is widely used in reinforcement learning with verifiable rewards for language-model reasoning. While value evolution along…
- DoSA: Dots Sparse Attention
Sparse attention visits few keys, yet finding them can still require scoring every key with every indexer head. We introduce DoSA (Dots Sparse Attention), a…
行业新闻
- AI agents aren't 'going rogue.' Frontier labs are skipping security basics
- Anthropic's Scion is heavily inspired by Mononoke
- TV4 postpones show after AI agent erases edited material
- Why do LLM pelicans look like this?
- Linus Torvalds's first LLM-assisted patch series
- Nvidia's first GPU: GeForce 256
- What agents mean for Rust (and vice versa)
- Ask HN: Why do people say "I built it" when we should say "Claude built it"?
It's just something I've been wondering about. Everyone says, “I built it,” but really we should say, “Claude built it for me,” or maybe, “I asked Claude to…
- I Deleted 168,000 Lines of Comments. My AI Agent Didn't Notice
- Anthropic revenue jumps to $16.8B in third quarter
社区动态
- We’ve opened an office in Singapore—our fifth in the Asia-Pacific region.
We’ve opened an office in Singapore—our fifth in the Asia-Pacific region. We’ve also signed an MOU with the Government of Singapore on AI safety and…
- Gemini 4 Argon is already unlocking new scientific breakthroughs in quantum computing, nuclear fusion, protein design…
Gemini 4 Argon is already unlocking new scientific breakthroughs in quantum computing, nuclear fusion, protein design and beyond. 🧵 https://t.co/Shj3Y4whoK
- European businesses are rapidly expanding their use of AI agents.
European businesses are rapidly expanding their use of AI agents. More than half of the AI use among our EU enterprise and business customers now involves…
- So many models launching this week!
So many models launching this week! And we have one of our own coming...⌛️ https://t.co/s5Z9QZ984W
- Remember alphaXiv? Their new research agents are told to report on users without telling them with no disclosure [D]
I was curious about OpenResearch by alphaXiv. "Turn your coding agents into research agents"? Sounds great! Just what I needed to save time on my experiments.…
- Gemini 4 Argon apps to try in AI Studio 🧵 1) Argon Tennis https://t.co/jdvu5Q82UF https://t.co/6Q9btrW8fr
Gemini 4 Argon apps to try in AI Studio 🧵 1) Argon Tennis https://t.co/jdvu5Q82UF https://t.co/6Q9btrW8fr
- Reasoning models burn tokens they don't need: a training-free contrastive decoding method shortens the chain of thought without retraining
- Textbook biology, brought to life right in your browser 🌿 This interactive 5-stage visual atlas built by Argon in…
Textbook biology, brought to life right in your browser 🌿 This interactive 5-stage visual atlas built by Argon in @antigravity turns dense textbook biology…
- cuda, hip: GDN chunked kernel by am17an · Pull Request #29353 · ggml-org/llama.cpp
Performance on Model Prompt tokens master tok/s pr tok/s Change RTX PRO 6000 Blackwell Qwen3.5-35B-A3B-Q8_0 2048 13,653.58 16,459.56 +20.6% RTX PRO 6000…
- Is the code you vibe secure? A 450-task benchmark built from 150 real vulnerabilities tests AI coding agents across 7 languages
官方发布
- Introducing Grok Voice 3.0
- mistralai/Timestral-123M
time-series-forecasting
- mistralai/Timestral-28M
time-series-forecasting
- mistralai/Timestral-5M
time-series-forecasting
- Gemini 4 Argon: Built to solve your hardest problems
We’re giving developers, enterprises, scientists, and consumers access to Gemini 4 Argon, our latest frontier model.
- Qwen/Qwen-Image-2.1-Turbo
text-to-image
- Redesigning source control at Anthropic
- In Claude Managed Agents, the web_search tool now follows your organization's web search domain restrictions set in…
In Claude Managed Agents, the web_search tool now follows your organization's web search domain restrictions set in the Claude Console. The agent receives…
- We've launched OpenTelemetry export in beta, so you can trace Messages API requests and Claude Managed Agents sessions…
We've launched OpenTelemetry export in beta, so you can trace Messages API requests and Claude Managed Agents sessions in your observability backend. The…
- Asana cuts model costs 76x in browser tests with GPT-6.1 Sol
Using GPT-6 Astra in Codex, Asana made its browser agent 76x cheaper and 5x faster in tests to offer customers more capable models.