Meta Debuts Muse Spark Open with Apple Tuning Alongside Research on AI's Gender Divide

Meta Debuts Muse Spark Open with Apple Tuning Alongside Research on AI's Gender Divide The Rundown newsletter highlights Meta's release of its 816B-parameter Muse Spark Open model configured to run across two Mac Studios, new Harvard findings uncovering significant AI gender gaps in workplace training and media exposure, and several industry updates from Amazon, OpenAI, and Anthropic. Meta released Muse Spark Open, an 816B-parameter open-weight model tuned with Apple to run locally on two M5 Ultra Mac Studios via MLX. Harvard researchers surveyed over 5,400 adults and examined 78,000 news articles, uncovering workplace training gaps and a 14.6-to-1 male bias in AI headlines. Amazon released two Bee AI wearables starting at $74.99 to capture audio, generate summaries, and integrate with Alexa. Anthropic granted six external METR researchers employee-level clearance to independently audit safety practices without approval barriers. Bank of America brought on 1,000 contractors to fix software flaws uncovered by Anthropic's Mythos and other models, while OpenAI trains staff against corporate espionage. Meta Releases Muse Spark Open with Apple Hardware Support Meta launched Muse Spark Open, an 816-billion-parameter open-weight multimodal model aimed at agentic tasks and coding. Through tuning with Apple, the model can execute locally across two M5 Ultra Mac Studios equipped with a combined 1TB of memory using Apple's MLX toolkit. Meta reported the model handled iOS application fixes and DaVinci Resolve video edits without cloud connectivity. On Artificial Analysis' Intelligence Index, Muse Spark Open scored 39, placing it ahead of Western open competitors but behind Chinese models including GLM 5.3, Kimi K3, and Qwen3.8, which maintain an edge in coding benchmarks. Harvard Study Maps AI Gender Gaps A study by Harvard Kennedy School's Shorenstein Center analyzed survey data from more than 5,400 adults, user browsing records, and 78,000 news stories to evaluate gender disparities in artificial intelligence. While browsing logs showed little difference in initial experimentation, 64% of employed women reported receiving no workplace AI training compared to 49% of men. Furthermore, news headlines referenced men 14.6 times more often than women, reflecting disparities in corporate leadership and quoted subject-matter experts. Industry Developments and Workflows Other major developments include Amazon's launch of two Bee wearable AI devices starting at $74.99 and Anthropic granting six METR researchers employee-level access to inspect safety without pre-publication approval. Bank of America hired 1,000 contractors to address vulnerabilities identified by Anthropic's Mythos, and OpenAI introduced anti-espionage training alongside GPT-Live-1.1 and an upgraded Chrome extension. Practical operational guides in the issue highlighted scheduled Claude agents for remote operational audits and prompt methods for ChatGPT's interactive interface.

Meta联手苹果推开源模型Muse Spark Open，哈佛发布AI性别差距报告 Meta发布816B多模态开源模型Muse Spark Open并支持双Mac Studio本地运行；哈佛研究揭示职场与报道中显著的AI性别差距；多家科技巨头同步更新AI硬件与安全动态。 Meta推出816B参数开源模型Muse Spark Open，联合苹果通过MLX实现在两台Mac Studio上本地运行 哈佛大学研究显示，虽然男女AI试用率相近，但在职场培训、使用信心及媒体曝光度上存在显著性别差异 亚马逊推出74.99美元起售的Bee系列AI可穿戴设备 OpenAI加强防范间谍与机密窃取培训，并推出GPT-Live-1.1语音AI及更新Chrome扩展 Anthropic向6位METR研究员开放内部权限独立评估安全；美银聘用千名外包修复因AI暴露的漏洞 Meta与苹果合作推进开源大模型本地化部署 Meta发布了8160亿参数的多模态开源模型Muse Spark Open，主打代码编写与智能体任务。借助苹果升级的MLX工具包，两台合计1TB统一内存的M5 Ultra Mac Studio即可在无需云端的情况下本地运行该模型，并成功执行了iOS应用修复和视频剪辑任务。在Artificial Analysis测试中该模型得分为39分，虽领先西方其他开源模型，但在代码能力上仍落后于GLM 5.3、Kimi K3和Qwen3.8等中国开源模型。 哈佛研究揭示AI领域的全方位性别分歧 哈佛大学肖伦斯坦中心基于5400多名成年人调查、浏览记录和7.8万篇新闻报道发现，AI领域的性别鸿沟主要体现在支持与声量上。尽管尝试AI的用户比例基本持平，但64%的受聘女性表示从未接受过AI培训（男性为49%）；在专业测试中女性更多选择“不确定”，而更多男性声称认识捏造的AI概念。此外，AI相关新闻标题中提及男性的频率是女性的14.6倍。 行业快讯与安全动态 各大厂商近期动态密集：亚马逊发布了起售价74.99美元的Bee AI穿戴设备，支持对话记录与总结；OpenAI发布改进多语言口音的GPT-Live-1.1，更新其浏览器扩展，并针对商业间谍活动对员工进行防范培训；Anthropic为第三方安全机构METR提供免审查的独立评估权限；美国银行则因Anthropic Mythos等模型暴露的漏洞，聘用了上千名承包商修复软件缺陷。

[Read Online](https://therundownai.beehiiv.com/p/meta-turns-mac-into-an-ai-supercomputer) ^| [Sign Up](https://www.therundown.ai/subscribe?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) ^| [Advertise](https://therundownai.typeform.com/to/kraZ1TSO?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) **Good morning, {{ first\_name | AI enthusiasts }}, and welcome to our 5,678 new readers.** Earlier this week, Meta caught up with OpenAI and Anthropic with its frontier model Muse Wonder. Turns out, the comeback story had one more chapter: a large open-weight release with an unlikely partner. The company just dropped Muse Spark Open, calling it the “most capable” open-weight AI model from the U.S., and tuned it with Apple so that two Macs can work like an AI supercomputer. **In today’s AI rundown:** Meta gives the West its best open-weight model Rowan’s Corner: My AI walks my company’s factory floor Turn a ChatGPT answer into a tool you can use Report: Harvard maps AI’s gender divide **LATEST DEVELOPMENTS** Image source: Mark Zuckerberg on X **The Rundown:** Meta just [released](https://research.meta.ai/blog/muse-spark-open?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) Muse Spark Open, an open-weight, 816B-parameter multimodal model built for coding and agentic work, and worked with Apple so that it could run locally on Mac Studios — a first for a model in this size class, Meta says. **The details:** Two M5 Ultra Mac Studios with 1TB combined memory can run Spark Open via Apple’s updated MLX toolkit, with Meta saying the model shipped iOS app fixes and edited video in DaVinci Resolve without the cloud. On Artificial Analysis’ Intelligence Index, it sits at 39, ahead of the West’s open-weight models but behind China’s GLM 5.3 (45), Kimi K3 (44), and Qwen3.8 (40). Spark Open either matches or tops open rivals on multimodal and knowledge tasks, but Chinese models still code better, with Kimi, GLM, and Qwen all ahead. It’s the third major Western open launch this month, after Reflection AI’s Beam and Mistral’s 1T-parameter Le Chonk, whose weights arrive Oct. 27. **Why it matters:** Meta built its AI rep on open Llama models before going closed. Now it’s playing both sides with Wonder at the frontier and Spark Open giving devs serious AI they can run on their hardware. China still leads the open-model race, but Meta’s return puts an American heavyweight back in contention. **The Rundown:** HubSpot’s new free “200+ AI-Powered Income Ideas” guide offers actionable strategies to turn artificial intelligence into your own personal revenue generator — unlocking a gateway to financial innovation in the digital age. **With this guide, you can:** Explore hundreds of revenue-generating ideas across industries with real-world applications Follow step-by-step instructions that make AI accessible to everyone Adopt cutting-edge strategies to stay ahead in today’s fast-paced market **Rowan:** At an in-person company, the CEO can walk the factory floor. You watch how the company actually runs, catch errors in real time, spot slow handoffs, and identify processes worth automating. The best operators do it religiously. The problem is that The Rundown is fully remote and has just over 20 people. There is no floor to walk. So every Friday, I do what I call a digital factory inspection: I pick random important things across the company (support queries, newsletter posts, specific finance items, etc.) and review each one with the same lens. What’s slow? Which handoffs could disappear? What should be automated? This week, I handed the entire prep to AI. A scheduled Claude task now runs every Friday before I wake up, and it: Picks the random areas to inspect (randomness matters so nothing gets gamed) Pre-gathers the samples, links, and data for each area into a report in my Notion Checks last week’s findings against Slack to verify they actually got fixed before I move on If you run a remote company (or any team, or honestly just your own workload), steal this. I put [the exact scheduled task prompt in our Workflow Hub](https://app.therundown.ai/community/posts/0ab88067-320e-49ee-b12c-fa87e2da7d39?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019), ready to adapt to your tools in about 15 minutes. It’s unglamorous plumbing, but it’s some of the highest-leverage 30 minutes of my week. **The Rundown:** In this guide, you’ll try ChatGPT’s new Intelligent UI by creating a workshop cost calculator. You’ll change the headcount, check the figures, and ask it to add a budget buffer. **Step-by-step:** Open ChatGPT in a web browser and start a Chat with GPT-6. Choose Medium where available and leave Pro reasoning off for this exercise Paste the guide’s prompt to create a calculator with editable costs, budget, and attendance. Check that the starting total is $700 for 20 people Change attendance to 30 using its controls. The total should rise to $950 and show that you are $50 over budget Try zero attendees and check that the fixed room fee stays. Test the budget boundary too, so a useful-looking calculator does not hide a wrong result Return to 20 people and ask for a 10% buffer. Check for a $770 final cost and $130 left, then change the headcount again to test the update **Pro tip:** Tell ChatGPT which inputs should change and which rules stay fixed. After each revision, check one result you know and one new case before using the tool. **The Rundown:** How are top organizations realizing value as AI spending accelerates? This webinar explores the leadership, operating model, and organizational shifts required to deliver measurable business and financial value with AI. **In this webinar, you’ll learn:** How to cut through the AI hype Why there is a gap between AI access and business impact What leading CEOs do differently to scale AI Image source: Harvard Kennedy School’s Shorenstein Center **The Rundown:** Harvard researchers just [found](https://shorensteincenter.org/resource/women-ai/?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) consistent gender gaps in how Americans use AI, get support for it, understand it, and appear in coverage of it, drawing on a survey of 5,400+ adults, browsing data, and 78K news articles. 29% of women said they’d never used AI vs. 22% of men, though browsing data showed no real gap in trying it, only in how much time users spent on it. At work, 64% of employed women reported no AI training vs. 49% of men, with the study linking training and encouragement to higher use and confidence for both. Men answered more AI questions correctly while women chose “not sure” more often, though more men also claimed to recognize made-up AI figures. AI headlines named men 14.6x as often as women, a gap the researchers partly tie to male-led companies, though the gap also held among quoted experts. **Why it matters:** This is one of the broadest looks yet at how AI’s gender gaps are taking shape. Access looks nearly equal today, but the real divide shows up in training, confidence, and who gets heard. As AI labs pitch superintelligence for everyone, these gaps show how far that promise still has to go. **QUICK HITS** **▸ Tony used AI agents to build and refine a word puzzle** Today’s workflow comes from reader **Tony Surfwell**: “I’m not a professional developer, but I built Sujido, a live word puzzle, with multiple AI agents. For months, they gave me bloat, lost context, and slop, and neither the models nor I could untangle what had been decided or why. I eventually settled on two agents: one Claude instance that plans and reviews, and another that builds. I imposed rules on both, one failure at a time. None of these rules would surprise a pro engineering team, but nobody handed them to me. That’s the lesson I learned: agents won’t bring the discipline. You have to impose it. When I showed the first game revision to strangers, they said that the first half was ‘boring’ and the second half was novel. It was hard news to hear about my baby, but I took the feedback and cut the game in half. That required surgery on code tangled across both halves, followed by a rebuild under the rules below. Before the rebuild, a separate AI audit graded the codebase B−. Afterward, it graded it A−. The site shrank by 97.5%, and every serious finding was fixed and locked in by an automated test. The models didn’t change drastically, but my operating rules did.” [See the workflow on our Community Workflow Hub →](https://app.therundown.ai/community/posts/4205e015-7e73-4337-aa34-4b262c5f5d77?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) **What have you fixed or built with AI lately?** **[Share your workflow](https://app.therundown.ai/community?submit=open&utm_source=therundownai.beehiiv.com&utm_medium=referral&utm_campaign=rundown-1012)**, and you could be featured right here in The Rundown. 🧠 [Muse Spark Open](https://huggingface.co/collections/meta-models/muse-spark-open?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) - Meta’s open-weight AI for coding and long-horizon agents 🗣️ [GPT-Live-1.1](https://x.com/OpenAI/status/2110797830361436604?s=20&utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) - OpenAI’s new voice AI with more natural multilingual accents 🤖 [Nemotron 3.5 Super VL](https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-Base-BF16?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) - Nvidia’s open multimodal AI for specialized AI agents 🖼️ [FLUX 3 Image Klein](https://huggingface.co/collections/black-forest-labs/flux-3-image-klein?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) - BFL’s small open model for 4K image generation, editing **Amazon’s Bee** [launched](https://www.aboutamazon.com/news/devices/amazon-personal-ai-bee-frontier-bee-eclipse?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) two AI wearables starting at $74.99 that record conversations, generate summaries and to-dos, and integrate with Alexa and productivity apps. **Anthropic** [said](https://www.anthropic.com/news/an-update-on-independent-evaluators?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) it is giving six METR researchers employee-level access to assess its AI safety practices and publish findings without its approval. **Bank of America CEO** Brian Moynihan [told](https://www.cnbc.com/2026/10/14/anthropic-mythos-bank-of-america-hiring-cost.html?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) CNBC that the bank hired 1K contractors to patch software flaws exposed by Anthropic’s Mythos and other models. **OpenAI** is [training](https://www.wsj.com/tech/ai/ai-companies-play-cat-and-mouse-with-secret-stealing-spies-160eec6b?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) employees to guard against espionage, including “honeypot” schemes, as attempts to steal AI trade secrets intensify, the Wall Street Journal reported. **OpenAI** also [launched](https://x.com/OpenAI/status/2110797830361436604?s=20&utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) GPT-Live-1.1 with improved multilingual accents and upgraded ChatGPT’s Chrome extension to let users analyze open tabs, automate browser tasks, and access chat history in the sidebar. Read our last AI newsletter: [Rogue AI agents sit for their first interview](https://www.therundown.ai/articles/rogue-ai-agents-sit-for-their-first-interview?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) Read our last Tech newsletter: [Apple goes all-in on the home](https://tech.therundown.ai/p/apple-goes-all-in-on-the-home?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) Read our last Robotics newsletter: [Inside Amazon’s million-robot experiment](https://robotnews.therundown.ai/p/inside-amazon-million-robot-experiment?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) Today’s AI tool guide: [Turn a ChatGPT answer into a tool you can use](https://app.therundown.ai/guides/chatgpt-intelligent-ui-guide?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) RSVP to next workshop on Oct. 21: [Get real work out of Microsoft Copilot](https://app.therundown.ai/live/getting-real-work-out-of-microsoft-copilot?utm_source=therundownai.beehiiv.com&utm_medium=newsletter&utm_campaign=rundown-1019) See you soon, *Rowan, Zach, Shubham, Jennifer, and Nate — the humans behind The Rundown*