오늘의 AI 소식

원제 — I read the GPT-6 Astra system card and I think we all misunderstand what "monitorability" means now

“I'm not a safety researcher, just build stuff with LLM's. Read some of the actual 30 page card, not the summary, and on…”

대상 주장 — Model dropped from 84.8% to 68.9%.

나온 때 2026-09-27

가져온 때 2026-09-28 05:05 UTC · 385b2606c5e1481d

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — 2026 in LLMs (so far)

“On Friday I gave the closing keynote at the WeAreDevelopers World Congress North America in San Jose. I tied together the key t…”

대상 주장 — # Another key trend in 2026 has been a dramatic improvement in the abilities of open weight models, including models that you can run on a laptop.

나온 때 2026-09-27

가져온 때 2026-09-28 05:05 UTC · 234343e34fa263e0

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — Anthropic’s CEO is about to have dinner with President Trump

“This will be the first one-on-one meeting between Dario Amodei and Donald Trump”

대상 주장 — Disrupt ticket savings of up to $200 + 50% off a second ends Sept 25, 11:59 p.m. PT.

나온 때 2026-09-27

가져온 때 2026-09-27 23:01 UTC · 747bb50a380ea03b

뜯은 자리 — 원문 페이지

원제 — Unsealed Briefs in Authors’ Case v. Microsoft/OpenAI

제목과 링크만 주는 소스라 인용할 것이 없다.

대상 주장 — OpenAI and Microsoft Knew Their Systems Would Replace Human Writers – OpenAI Policy Director Jack Clark stated in May 2020: “Our work on AI and Creativity is going to increasingly lea…

나온 때 2026-09-27

가져온 때 2026-09-27 23:01 UTC · 88ff5154f4eecae4

뜯은 자리 — 원문 페이지

원제 — OpenRouter: from Seed to Stripe — with OpenRouter’s Alex Atallah & AMP’s Anjney Midha

“From the earliest days of open-weight models to becoming the neutral routing layer for more than 10 million developers, OpenRou…”

대상 주장 — ” It outperforms 3 on, one or two benchmarks.

나온 때 2026-09-25

가져온 때 2026-09-26 23:01 UTC · dd0fdd232648b2a0

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — Revealing the details of how OpenAI agents hacked Hugging Face

제목과 링크만 주는 소스라 인용할 것이 없다.

대상 주장 — Over the subsequent two weeks, we wrote programs to follow the chains and decoded over 80,000 payloads.

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · c30cd0f582019bca

뜯은 자리 — 원문 페이지

원제 — Proaction boosts sales 60% and saves 75+ hours with Codex

“With Codex, GPT-Live-1, and GPT-6 Astra, Proaction builds, operates, and sells modern fleet management faster.”

대상 주장 — With Codex, GPT-Live-1, and GPT-6 Astra, Proaction builds, operates, and sells modern fleet management faster.

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · 794516609b5bcb0e

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — Scaling MoE reinforcement learning on Amazon EKS with EFA and DeepEP with 40% more throughput

“Learn how to scale Mixture-of-Experts (MoE) reinforcement learning on Amazon EKS using Elastic Fabric Adapter (EFA) and DeepEP.…”

대상 주장 — Performance and cost optimizations This section covers two key optimizations: using DeepEP to reduce expert-parallel communication overhead over EFA, and using Amazon Elastic Compute Clou…

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · b9cac234d7f78a36

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — Accelerate multimodal RL training with SkyRL on Amazon SageMaker HyperPod

“Learn how to run SkyRL, an open-source reinforcement learning framework, on Amazon SageMaker HyperPod to post-train a Qwen3-VL-…”

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · a4bd252f1f6e25fa

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — NarrateAI: production-ready LLM quality assurance on Amazon Bedrock

“NarrateAI delivers production-ready LLM quality assurance on Amazon Bedrock. This post details five techniques—adaptive pipel…”

대상 주장 — NarrateAI transforms business intelligence for over 4,000 AWS executive leaders through a two-layer architecture built on Amazon Bedrock AgentCore, a platform to build, connect, and optim…

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · 4c2843cb765a0255

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — Deploying real-time personalized speech with Qwen3-TTS on Amazon SageMaker AI

“Deploy the publicly available Qwen3-TTS-12Hz-1.7B-Base text-to-speech model from Amazon SageMaker JumpStart to a fully managed,…”

대상 주장 — g6.4xlarge (1x NVIDIA L4 GPU, 24 GB), which is sufficient for the 1.7B model.

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · c28aa6c37d511074

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — How Datacor built self-service rental analytics with Amazon Quick Sight

“Learn how Datacor built a self-service rental analytics experience for gas and welding distributors by embedding Amazon Quick S…”

대상 주장 — An Azure Data Factory pipeline extracts source tables as Apache Parquet files and streams them into Amazon Simple Storage Service (Amazon S3).

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · 09c0269709886573

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — Multi-Region training with Amazon SageMaker HyperPod and Qumulo

“Amazon SageMaker HyperPod and Cloud Native Qumulo let you place training compute in one AWS Region while keeping your dataset i…”

대상 주장 — The spoke cluster runs in the US West (Oregon) Region (us-west-2) and reads the data remotely through CDF with 60 ms of network latency.

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · ad552593d4bb8e33

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — U.S. appeals court upholds designation of Anthropic as supply chain risk

제목과 링크만 주는 소스라 인용할 것이 없다.

대상 주장 — BreakingNews-stickyContainer):before{animation:BreakingNews-SlideInFull 2s ease;animation-iteration-count:1;background:linear-gradient(90deg,#fff0 0,#fc3c46 30%,#fc3c46 70%,#fff0);backgro…

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · aa428a07a7ec26c1

뜯은 자리 — 원문 페이지

원제 — Runway’s WorldPrompt and the Engineering of Real-Time Worlds

“Earlier this month, world model company Runway introduced GWM Worlds 2, a research preview that “turns high-fidelity video an…”

대상 주장 — Alongside Runway, there are several other notable projects in this domain: Google DeepMind’s Genie 3 (which also generates at 720p and 24 fps), Odyssey-2 Pro, and World Labs’ RTFM (Re…

나온 때 2026-09-25

가져온 때 2026-09-25 23:01 UTC · 6a41c2bc2e748668

뜯은 자리 — 목록뿐 (원문을 못 읽었다)

원제 — FuseReg: Regularizing Layer Fusion Mitigates the Reconstruction-Generation Gap in Representation Autoencoders

“Representation autoencoders (RAEs) reuse features from a pretrained visual encoder as reconstruction and diffusion latents, int…”

대상 주장 — On ImageNet-256 with DINOv3-L, a single FuseReg decoder reconstructs from full, sparse, and single-layer fusions without retraining, achieving higher PSNR than decoders specialized to fix…

나온 때 2026-09-24

가져온 때 2026-09-28 05:05 UTC · 6861a4d455b6989e

뜯은 자리 — 초록

원제 — InternW0-Δ: A World Action Model Bridging Predictive Dynamics and Actions with 20K+ Hours of Open Data

“World Action Models (WAMs) jointly model visual dynamics and action generation for generalist robot manipulation. A central cha…”

대상 주장 — We introduce InternW0-Δ, a unified WAM pretrained on a heterogeneous corpus that outperforms prior methods across simulation benchmarks and real-robot platforms.

나온 때 2026-09-24

가져온 때 2026-09-28 05:05 UTC · fce8fc3803ad1364

뜯은 자리 — 초록

원제 — Block Sparse Attention with Log-Linear Complexity

“Scaling language models to long contexts is limited by the quadratic cost of self-attention. Block sparse attention offers an e…”

나온 때 2026-09-24

가져온 때 2026-09-28 05:05 UTC · e8ca90729f4aeb76

뜯은 자리 — 초록

원제 — orcarouter/OrcaSAQ-2-27B

제목과 링크만 주는 소스라 인용할 것이 없다.

대상 주장 — OrcaSAQ2 performs strongly on long-horizon workloads relative to models in its deployment and parameter class, despite operating from a 12.3 GB checkpoint.

나온 때 2026-09-24

가져온 때 2026-09-26 23:01 UTC · 82644ada59f66f78

뜯은 자리 — 모델 카드

원제 — Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents

“The rapid progression of large language models is extending AI from passive content generation into the active workflows of eng…”

나온 때 2026-09-23

가져온 때 2026-09-25 23:01 UTC · 4f969e088a867b41

뜯은 자리 — 초록

원제 — Coding Agents for Generalized Task and Motion Planning Problems

“Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discre…”

대상 주장 — Across all program synthesis methods, we evaluate 980 generated programs on 100 held-out instances each, 98,000 evaluation episodes in total.

나온 때 2026-09-23

가져온 때 2026-09-25 23:01 UTC · 8f9867a434213048

뜯은 자리 — 초록

원제 — ExplorationBench: Measuring AI Systems' Exploration in Verifiable Alien Worlds

“Scientific discovery begins where known problems end. There, AI systems must engage in exploration: framing hypotheses, designi…”

대상 주장 — We evaluate 10 AI systems and find that the strongest systems can acquire and apply unfamiliar rules, while performance varies substantially across trajectories and continued exploration …

나온 때 2026-09-23

가져온 때 2026-09-25 23:01 UTC · 9ccf356c7f3c30e0

뜯은 자리 — 초록

원제 — WanPE: Towards Cinematic Prompt Enhancement for Modern Text-to-Video Generation

“Video generation begins in text space by authoring a cinematic screenplay, then materializes into pixels. As contemporary video…”

나온 때 2026-09-23

가져온 때 2026-09-25 23:01 UTC · 01decd428e012e3f

뜯은 자리 — 초록

원제 — Rufus-Air: An Open LLM Post-Training Recipe

“Rufus-Air is an open and reproducible post-training recipe on GLM-4.5-Air-Base (106B-A12B), organized as a serial pipeline of e…”

대상 주장 — Rufus-Air improves over the official GLM-4.5-Air post-trained release and is competitive with similarly sized open models.

나온 때 2026-09-23

가져온 때 2026-09-25 23:01 UTC · ecf373efc650dcaf

뜯은 자리 — 초록

원제 — Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs

“While Large Language Models (LLMs) rely on highly non-linear components, in this work we demonstrate that they exhibit fundamen…”

나온 때 2026-09-23

가져온 때 2026-09-25 23:01 UTC · 859a30c4317d1a6d

뜯은 자리 — 초록

원제 — yetone/magpie

“Every agent's model. One place. Codex on DeepSeek, Claude Code on Kimi, from the menu bar.”

나온 때 2026-09-23

가져온 때 2026-09-26 23:01 UTC · 7abcc96641ec3723

계정이 적어 둔 문자열 — China

뜯은 자리 — README

원제 — SupersonicLabs/Julia-1

제목과 링크만 주는 소스라 인용할 것이 없다.

대상 주장 — Typed accuracy: Choice 71.33% (428/600), Noul 80.67% (484/600), Score 68.88% (551/800).

나온 때 2026-09-23

가져온 때 2026-09-27 23:01 UTC · f83624fe1b9581b4

뜯은 자리 — 모델 카드

원제 — fastino/GLiNER2.5-Decide

제목과 링크만 주는 소스라 인용할 것이 없다.

나온 때 2026-09-23

가져온 때 2026-09-26 23:01 UTC · b267fd54205b20a2

뜯은 자리 — 모델 카드

원제 — JohnHeibel/PDoomVideo

“Source code for the Claude Opus 5.5 music video for I'm Upping My P(doom)”

나온 때 2026-09-22

가져온 때 2026-09-25 23:01 UTC · 683af84eb2b0acdc

계정이 적어 둔 문자열 — Berkeley, California

뜯은 자리 — README

원제 — Learning to Discover Interesting Mathematics

“Recently, Large Language Models (LLMs) have been increasingly able to solve advanced mathematical problems, including many that…”

대상 주장 — Optimizing for our metric creates a model capable of producing more interesting theorems, while also reducing substantial or full overlap with Mathlib from 91.9% to 30.6%, showcasing the …

나온 때 2026-09-22

가져온 때 2026-09-25 23:01 UTC · 6d69919b7b03eed9

뜯은 자리 — 초록