#01
Fay：fay是一个帮助数字人（2.5d、3d、移动、pc、网页）或大语言模型（openai兼容、deepseek）连通业务系统的agent框架。
fay是一个帮助数字人（2.5d、3d、移动、pc、网页）或大语言模型（openai兼容、deepseek）连通业务系统的agent框架。
推荐理由数字人 Agent 框架 · 13482 star
周报
2026-W32 周报精选：Fay：fay是一个帮助数字人（2.5d、3d、移动、p；How Zapier transformed core ；Meta is back with Muse Glimm
fay是一个帮助数字人（2.5d、3d、移动、pc、网页）或大语言模型（openai兼容、deepseek）连通业务系统的agent框架。
The enterprise marketing team at Zapier uses ChatGPT Work to reduce the number of drop-offs in its lead funnel, build campaign assets, and automate reporting.
Learn how new AI and agentic experiences across Google Ads and Google Analytics can simplify your marketing workflow.
OpenAI sent Governor Greg Abbott a letter outlining its commitment to responsible AI infrastructure in Texas. The letter supports reliable, transparent growth that benefits Texans.
Model ML uses GPT-5.6 Sol to carry finance work from research and analysis through editable, traceable PowerPoint decks and Excel workbooks.
Meet GPT-5.6-Cyber, OpenAI’s cybersecurity-specific model available through Daybreak Red for authorized vulnerability research, exploit validation, and security testing.
Approved Daybreak partners can use OpenAI’s frontier cyber models to deliver authorized, governed cybersecurity services to customers.
Virgin Atlantic is accelerating research, product planning, and decision-making with ChatGPT Work, helping teams connect signals across the customer journey.
Premium seats are now available on ChatGPT Business, with 5x more usage, no five-hour usage limit, and flexible seat options for every teammate.
Musings on model alignment, what determines safety, and where we go from here.
Discover how HSP GRUPPE uses ChatGPT Enterprise to boost productivity, improve work quality, and create more capacity for tax advisory and client service.
OpenAI is sharing preliminary cybersecurity evaluations for Astra and the steps we’re taking to strengthen safeguards and security controls.
**OpenAI** escalates its upcoming **Astra** model to "critical" cyber status due to significant advancements in agentic coding and cybersecurity, pausing some activities to strengthen controls. The "Hugging Face incident
Meta returns to the open source ecosystem with the release of Muse Glimmer, a 30B open-weight dense model with a 120K+ context window built for local AI...
Meta's Muse Glimmer, the first open model released by Meta Superintelligence Labs, is now available. Muse Glimmer is a 30B multimodal model released under the Apache 2.0 license, designed for local coding agents, and acc
The post How to build an AI agent: A simple guide for anyone appeared first on Source .
Improving Fable 5's biology safeguards
ChatGPT introduces improved GPT-5.6 Sol with better accuracy and consistency, plus expanded access for free users and unlimited everyday chats with GPT-5.6 Luna.
OpenAI and the American Psychological Association advance evidence-based guidance, resources, and safeguards for responsible AI use and youth mental health.
New OpenAI Signals data shows how people use ChatGPT worldwide, with country-level insights on adoption, usage trends, and evolving behavior.
**Meta's Muse Spark 1.2** rapidly rose to frontier-tier with top 5 ranking on Vals Index at **$0.69/test**, being **3x cheaper than Kimi** and **10x+ cheaper than Fable, Opus, and 5.6 Sol**. It achieved **gold-medal-leve
OpenAI explains recent third-party cybersecurity evaluation incidents and outlines new safeguards to strengthen AI model testing and evaluation.
We ran 900 DeepSWE rollouts on DeepSeek-V4 Flash and GPT-5.6 Luna. Luna leads pass@1 by 14 points; DeepSeek delivers 4.8x the solves per dollar.
Explore new education plugins for ChatGPT Work and Codex that help K–12 teachers, college educators, and students learn, teach, research, and build.
Kaggle’s AI Agents Intensive with Google brought learners together in a no-cost course to build and deploy the next frontier of AI.
GPT-Live enables continuous voice interaction with AI, using a turnless speech model and low-latency architecture for faster, more natural conversations.
**Google DeepMind** undergoes a leadership reshuffle with **Demis Hassabis** moving to Chair and Chief Scientist roles, while **Koray Kavukcuoglu** takes operational control focusing on **Gemini** and product execution.
OpenAI addresses Apple’s baseless lawsuit, corrects claims about its employees, and shares messages documenting what happened.
Circles uses the OpenAI API and Codex to power AI-native telco experiences, increasing ARPU by 22%, reducing churn by 9%, and improving development efficiency.
Here are Google’s latest AI updates from July 2026
Dan Shipper 上个月的 Codex 账单是 13,000 美元，他的 COO 给了他一个白眼。 但他不打算减少用量。 Every 的 Slack 里现在有 20 个 agent 和 20 个人类，一起工作。 代码、写作、客服、邮件、研究备忘录，能用 AI 处理的流程全部交给了 agent。 27 个全职员工，
Mariano-Florentino (Tino) Cuéllar to join Anthropic as Chief Global Affairs Officer
Shieldstral introduces a 3B open-weights multimodal safety classifier that outperforms models up to 7x its size.
See how AI enriches employee experience: Frontier organizations help people grow, do meaningful work, and thrive, beyond time saved. The post Accelerating Frontier Transformation: Enrich employee experience appeared firs
Scaling our curation and measurement of the open ecosystem.
https://www.xiaoyuzhoufm.com/episode/6a69b07eb581962ce2bd4d97 2024 年 4 月，一篇挂在 arXiv 上的论文在技术社群里炸开了锅。 有人说，统治深度学习几十年的 MLP（多层感知机）可能要被改写了。 也有人说，这不过是又一个"看起来很美"的架构。 这篇
When do we build the moon arcology?
Storage is an active part of every agentic AI workflow. As agents retrieve enterprise knowledge, access persistent memory, reuse key-value (KV) cache data,...
The post Teaching AI to speak the language of pathology appeared first on Source .
**Alibaba** launched **Qwen3.8-Max**, a **2.4T-parameter** open-weight model emphasizing autonomous coding, long-horizon execution, and multimodal feedback, with aggressive pricing. Early benchmarks rank it highly on hum
Anthropic releases Opus 5 promising Fable 5-like capabilities, Google Releases Three New Gemini A.I. Models, and more!
这期商业访谈录也超级精彩，了解到很多关于 Deepseek 的厉害之处。 小宇宙博客地址： https://www.xiaoyuzhoufm.com/episode/6a66ed17a3fec224d5a3f744 AI 重写成文章，帮没空听的朋友节省3小时： 有人在公司注册前夕接到电话，对方开门见山：你们四个创始人，
**Alibaba** launched **Qwen3.8-Max**, enhancing multimodal capabilities and agent ecosystem integration. **NVIDIA** introduced **Alpamayo 2 Super** for autonomous vehicle reasoning, while **Mistral AI** released **Shield
1930年，一位英国语言学家出版了一本书，声称用850个英语单词就能完成日常生活的全部表达。 这听起来不可思议。 英语词汇量超过17万，莎士比亚一个人就用了两万多个不同单词。 850个，够干什么？ 但这个想法在二战结束后引发了全球范围的讨论，影响了BBC的广播方式，塑造了今天全球英语教学的基础词汇体系，甚至间接启发了乔
Sam Altman breaks down the debate around AI infrastructure. He addresses why people worry about data centers in their towns, how closed-loop cooling fixed massive water usage, and why AI supercomputing hubs are transitio
**Frontier API vulnerability** revealed exposure of hidden reasoning traces including sensitive data like **62 unique API keys** and **33 passwords**, raising privacy and operational-security concerns. Discussions highli
World-Action Models (WAMs) improve end-to-end autonomous driving by transferring video dynamics priors to action prediction, but existing methods require costly future generation at inference. We present SimWAM, a simple
From building agent-powered web applications to context-aware AI features, see how you can use CopilotKit integration for Angular in your workflows. Resources: AG-UI user interface for ADK → https://goo.gle/4yUSZFj
Sam Altman reveals a terrifying AI security incident where an unreleased model chained together multiple zero-day exploits to break out of its sandbox, access the internet, and hack Hugging Face to cheat on its evaluatio
Caroline di Vittorio, Engineer at LangChain, builds a voice agent with the Google ADK and the Gemini Live model, then sets up tracing in LangSmith to see exactly what the agent is doing under the hood. Gemini Live is Goo
From routing a 200,000-token prompt across GPUs to having GLM-5.2 profile, rewrite, and optimize the kernels serving itself, inference engineering is becoming one of the most important layers of AI. In this episode, Base
A world model is only useful for physical AI if it changes what the agent does, and only safe if it declines to do so when it is wrong. We study both halves of that requirement with CausalNav, a controller built around a
Large language models perform strongly on medical knowledge benchmarks, but reliable clinical deployment requires agents to conduct defensible investigations over heterogeneous, longitudinal records: determining what evi
Recent advances in generative recommendation (GR) leverage large language models (LLMs) as recommender backbones, enabling LLMs to directly generate recommendations conditioned on item-interaction histories. In these sys
After a few long years of finding time to document my lessons from training open models, my post-training book is done!
Which galaxy will you choose?
**Meta** re-enters the open-weight frontier with the release of **Muse Glimmer**, a **30B dense**, multimodal, agent-focused model under **Apache 2.0**, optimized for always-on local agents and consumer hardware. It feat
An LLM judge deployed inside a reasoning pipeline does not merely measure quality, it decides which answer ships. We show that the cost of that decision depends less on judge accuracy than on the decision rule the judge
Gaussian-splatting proxies enable interactive rendering of volumetric medical scans, but a clipping plane exposes anatomy not constrained by external-view training and intersects primitives that conventional splatting ca
Cardiovascular AI models can classify clean elec- trocardiogram (ECG) signals, but real wearable signals change because of motion, breathing, posture, sensor contact, and true clinical deterioration. This paper asks when
Open-weight large language models (LLMs) are increasingly developed through complex, multi-stage pipelines, leading to intricate lineage relationships that reflect model origin, ownership, and evolution. Understanding th
Reinforcement learning (RL) with verifiable rewards constructs trajectory-level advantage estimates, yet it often fails to credit the few pivotal decisions that determine outcomes in long-horizon, multi-turn agentic task
Four years after prompt injection was first identified in 2022, attacks are still predominantly documented as verbatim strings rather than structured exploits, despite advancing agent capabilities and threat actors embed
Generating large-scale, freely explorable 3D worlds from open-ended text remains challenging because a system must jointly maintain global spatial coherence, rich local content, and explicit assets suitable for downstrea
LLM recommenders for top-K item suggestion regularly emit titles outside the target catalog. Prior audits report a binary out-of-domain rate; none ask whether the model knew. We jointly audit hallucination rate (OOD@10)
Computer-using agents (CUAs) are advancing rapidly across the digital world. A CUA trajectory records the agent's actions, states, and reasoning. Verifying whether it fulfilled the task instruction is central to CUA eval
Short segments of perceived speech can be retrieved from non-invasive magnetoencephalographic (MEG) recordings by deep networks trained with a CLIP-style objective against wav2vec 2.0 audio embeddings. Yet their weights
LLM benchmarks can build an organization's reputation and attract customers, but only when results are transparent and verifiable. Unverified claims that DeepSeek R1 outperformed OpenAI's o1 contributed to market panic o
Training large language model agents for long-horizon tool use typically relies on interactions with real or synthesized executable environments, whose construction and verification are costly, or on external simulators
Artificial intelligence is changing the task composition of computing work faster than curricula and training typically adapt. This is a curriculum-framework paper, grounded in a structured narrative review of labor-mark
Mobile agents have achieved promising results on clean online benchmarks such as AndroidWorld, yet their performance often degrades sharply in real-world deployment due to environmental variations and imperfect interface
As LLMs are increasingly deployed within agentic systems, their capabilities depend not only on the model weights but also on the harness: the prompts, tools, control flow, memory, and orchestration code surrounding them
Economic World Models (EWMs) are generative economic models that simulate how economies evolve from within by modeling heterogeneous agents, their beliefs and actions, and the market and institutional mechanisms through
Explanations play a crucial role in creating trustworthy recommender systems (RS), yet choosing a good explanation method presents challenges. Many explanation methods exist, but little guidance exists on which is best f
Spatial intelligence is fundamental to embodied agents, yet existing benchmarks focus on local spatial perception from single or few viewpoints, overlooking global spatial awareness over continuous, long-horizon visual s