Meta is opening up its Muse AI agent to hardware makers through a new open-source project called Muse Gadgets. The release includes ESP32 firmware, a Linux SDK, and developer documentation available under the Apache 2.0 license. Developers can run the tools on off-the-shelf boards like ESP32 or Raspberry Pi and add screens, buttons, sensors, and audio to build custom peripherals. Meta also shipped its own Muse Home Link device, a USB-C powered gadget that connects Muse to home networks so the AI agent can control smart TVs and speakers. The company is giving away 5,000 Home Link units to active U.S. subscribers for free shipping.
On the AI model side, Meta said Muse Spark helped mathematicians solve six open problems in areas like probability and differential equations. The models, Muse Spark 1.1 and 1.2, were accessed through the regular Meta AI chat interface. Meta also launched a Discord channel for Muse Gadgets users to share projects and get support.
In chip stock news, Intel and Marvell both sell AI-related silicon but follow very different paths. Marvell's fiscal 2026 revenue grew roughly 42 percent, with data center products making up 74 percent of sales. Intel, meanwhile, is building out a foundry business but still loses money on it and has issued large amounts of new stock. Analysts view Marvell as the stronger buy because it grows faster without operating its own factories.
NVIDIA and SK Hynix sit in different but connected parts of the AI hardware stack. NVIDIA designs the GPUs that power AI data centers and reported fiscal 2026 revenue of $215.9 billion with net income of $120.1 billion. SK Hynix makes the high-performance memory those systems need and posted fiscal 2025 revenue of $71.5 billion and net income of $31.6 billion. SK Hynix trades at lower valuation multiples, making it more value-oriented, while NVIDIA offers stronger growth and profit margins.
Micron is riding similar demand waves. The company reported past-year sales of $133.188 billion and net income of $84.969 billion, gave bullish fiscal 2027 guidance, and kept a quarterly dividend of $0.15 per share. More than 75 percent of 2027 output is already reserved under long-term customer deals, with AI demand for high bandwidth memory driving the growth.
IBM is making its agentic software development platform Bob available in self-hosted deployments. Enterprises can now run Bob on premises, in private clouds, or in air-gapped networks. Customers must supply their own supported model, such as NVIDIA Nemotron or Poolside Laguna, for full isolation. IBM also offers hybrid modes using Claude, Gemini, or GPT models, plus optional Premium Packages for Java, IBM i, and IBM Z modernization.
Alphabet's Google signed a White House accord on voluntary AI safety standards, with CEO Sundar Pichai endorsing the agreement. The pact calls for stronger AI security testing, clearer system labeling, and more transparent reporting. Google also won two antitrust lawsuits in U.S. federal courts, easing some legal pressure on parts of its business model.
On the regulatory and market front, Robinhood has separated its AI technology from its brokerage operations in a way that may help it avoid liability, a structure other brokers have not copied. Consulting firms warn that as AI becomes cheaper, power demand could keep rising and strain electricity supplies. The Bank for International Settlements studied 1,246 AI firms from 2021 to 2025 and found that 55.2 percent of investment came from other AI companies, raising concerns about circular funding and higher risks during market stress.
Cimpress raised its fiscal 2028 adjusted EBITDA target to at least $615 million, up from $600 million. The company reported fiscal 2026 revenue of $3.7 billion and adjusted EBITDA of $458 million, focusing on higher-value small-business customers and elevated products like packaging, apparel, and promotional items. Cimpress also sees generative AI as a growth driver across its Mass Customization Platform. Separately, Capcom plans to evolve its RE Engine into what it calls an AI-generation game engine through the REX Project, aiming to support over 2,000 developers and deeply integrate AI into game creation using tools like RE:Dox, RE:UI, and RE:Flows.
Key Takeaways
- Meta released Muse Gadgets, an open-source ESP32 firmware and Linux SDK for building custom hardware connected to its Muse AI agent, available under the Apache 2.0 license.
- Meta is giving away 5,000 Muse Home Link USB-C devices to active U.S. subscribers, letting the AI agent join home networks and control smart TVs and speakers.
- Muse Spark AI models helped mathematicians solve six open problems in probability and differential equations through the regular Meta AI chat interface.
- NVIDIA reported fiscal 2026 revenue of $215.9 billion and net income of $120.1 billion, while SK Hynix posted fiscal 2025 revenue of $71.5 billion and net income of $31.6 billion.
- Micron reported past-year sales of $133.188 billion and net income of $84.969 billion, with over 75 percent of 2027 output reserved under long-term customer deals.
- IBM made self-hosted deployment of its Bob agentic development platform generally available, supporting NVIDIA Nemotron, Poolside Laguna, Claude, Gemini, and GPT models.
- Alphabet CEO Sundar Pichai signed a White House AI safety accord calling for stronger security testing, clearer system labeling, and transparent reporting.
- Robinhood separated its AI technology from brokerage operations in a way that may help it avoid liability, a structure other brokers have not yet copied.
- The Bank for International Settlements found that 55.2 percent of investment in AI companies from 2021 to 2025 came from other AI companies, raising concerns about circular funding.
- Cimpress raised its fiscal 2028 adjusted EBITDA target to at least $615 million and sees generative AI as a growth driver across its Mass Customization Platform.
Meta launches Muse Gadgets open-source AI hardware tools
Meta released Muse Gadgets, an open-source project that lets developers build custom hardware for its Muse AI agent. The project includes ESP32 firmware, a Linux SDK, a GitHub repository, and developer documentation. Users can obtain API tokens from gadgets.muse.ai and use off-the-shelf boards like ESP32 or Raspberry Pi. The tools allow makers to add screens, buttons, sensors, and audio to Muse.
Meta unveils Muse Gadgets with open-source firmware and SDK
Meta introduced Muse Gadgets, an open-source project that lets developers and hobbyists connect the Muse AI agent to custom hardware. The project provides firmware and a Linux software development kit. It can run on Raspberry Pi computers and ESP32 boards. Meta also created a Discord channel for users. The company tested the concept with its own Muse Home Link device.
Meta Muse Spark solves six math problems, releases hardware tools
Meta said its Muse Spark AI models helped mathematicians solve six open problems in areas like probability and differential equations. The company also announced Muse Gadgets, an open-source ESP32 firmware and Linux SDK for building hardware that works with Muse. Muse Spark 1.1 and 1.2 were used through the regular Meta AI chat interface. Meta also released its own Muse Home Link gadget.
Meta open sources Muse Gadgets for developer hardware projects
Meta announced Muse Gadgets, an open source ESP32 firmware and Linux SDK released under the Apache 2.0 license. Developers can build peripherals using off-the-shelf microcontroller boards or Raspberry Pi machines. Meta also built Muse Home Link, a USB-C powered device that connects Muse to home networks to control smart TVs and speakers. Meta manufactured 5,000 Home Link units for free shipping to subscribers.
Meta opens Muse AI to DIY hardware makers with free devices
Meta launched Muse Gadgets, an open-source project providing firmware and a Linux SDK for custom hardware connected to its Muse AI agent. The company is giving away 5,000 Muse Home Link USB-C devices to active U.S. subscribers. Muse Home Link lets the AI agent join a home network and communicate with smart devices. The launch coincided with Apple announcing tighter macOS Full Disk Access controls for AI agents.
Intel vs Marvell: Which AI Chip Stock Is the Better Buy in 2026
Intel and Marvell Technology both sell AI-related chips but have very different business models. Marvell's fiscal 2026 revenue grew about 42 percent and data center products made up 74 percent of sales. Intel is building a foundry business but still loses money on it and has issued large amounts of new stock. The author views Marvell as the better buy because it grows faster without running its own factories.
NVIDIA vs SK Hynix: Which AI Stock Fits Your Portfolio
NVIDIA designs the powerful GPUs used in AI data centers, while SK Hynix makes the high-performance memory chips those systems need. NVIDIA reported fiscal 2026 revenue of $215.9 billion and net income of $120.1 billion. SK Hynix posted fiscal 2025 revenue of $71.5 billion and net income of $31.6 billion. SK Hynix looks more value-oriented based on lower valuation multiples, while NVIDIA offers stronger growth and profit margins.
Cimpress Aims for $615 Million EBITDA by 2028
Cimpress set a new target of at least $615 million in adjusted EBITDA by fiscal 2028, up from its prior goal of $600 million. The company reported fiscal 2026 revenue of $3.7 billion and adjusted EBITDA of $458 million. It is focusing on higher-value small-business customers and what it calls elevated products like packaging, apparel and promotional items. Cimpress also sees generative AI as a growth driver across its Mass Customization Platform.
AI Trading Platforms and the Corporate Liability Shield
Robinhood has separated its AI technology from its brokerage operations in a way that may help it avoid liability. The article describes this as a corporate shell game that other brokers have not yet copied. It also points out that the regulatory gap allowing this structure may not last long. The piece raises concerns about accountability in AI-driven trading platforms.
IBM Offers Self-Hosted Version of Bob AI Development Platform
IBM made a self-hosted deployment option for its agentic software development platform Bob generally available. Enterprises can now run Bob on premises, in private clouds or in air-gapped networks. Customers must supply their own supported model, such as NVIDIA Nemotron or Poolside Laguna for full isolation. IBM also offers hybrid modes using Claude, Gemini or GPT models and optional Premium Packages for Java, IBM i and IBM Z modernization.
AI cost drop may push power grid to limit
Consulting firms warn that as AI becomes cheaper, power demand could keep rising. Lower AI costs may lead more people and businesses to use AI. This growing use could strain electricity supplies. The trend points to a possible clash between tech growth and grid capacity.
Micron sees strong AI memory demand, sets dividend
Micron Technology reported past-year sales of US$133.188b and net income of US$84.969b. The company gave bullish fiscal 2027 guidance and kept a quarterly dividend of US$0.15 per share. More than 75% of 2027 output is already reserved under long-term customer deals. Management said AI demand for high bandwidth memory is driving the growth.
AI firms fund over half of AI company investments
The Bank for International Settlements studied 1,246 AI firms from 2021 to 2025. It found that 55.2% of investment in AI companies came from other AI companies. The report says this circular funding makes reported demand look stronger than real customer demand. It also raises concerns about higher risks during market stress.
Capcom turns RE Engine into AI-generation game engine
Capcom plans to evolve its RE Engine into what it calls an AI-generation game engine through the REX Project. The engine has been used since 2017 and powers about 30 Capcom titles. The update aims to speed up development, support over 2,000 developers and deeply build AI into game creation. Tools like RE:Dox, RE:UI and RE:Flows will help teams create games together with AI.
Alphabet signs AI safety pact and wins antitrust cases
Alphabet CEO Sundar Pichai signed a White House accord on voluntary AI safety standards. The agreement calls for stronger AI security testing, clearer system labeling and more transparent reporting. Google also won two antitrust lawsuits in US federal courts. These moves ease some legal pressure on parts of Alphabet's business model.
Sources
- Meta Open-Sources 'Muse Gadgets' To Let You Build Your Own AI Hardware: Here's How
- Meta introduces Muse Gadgets with open-source tools for custom hardware
- Meta says Muse Spark helped solve 6 major math problems, releases open-source project for AI hardware
- Meta Open Sources Muse Gadgets, Lets Developers Build Their Own Hardware For Its AI Agent
- Meta Opens Muse AI Agent to DIY Hardware Makers With Free Smart Home Devices
- Intel vs. Marvell Technology: Which AI Chip Stock Is a Better Buy in 2026?
- Better Artificial Intelligence Stock: NVIDIA vs. SK Hynix
- Cimpress Targets $615M EBITDA by 2028 as AI, Premium Products Fuel Growth
- The Corporate Shell Game: How AI Trading Platforms Are Building the Liability Escape Hatch
- IBM Brings Bob to Self-Hosted and Air-Gapped Environments: Agentic Software Development Without Moving Your Code
- Cheaper AI models may strain power grid, consulting firms say
- How Investors May Respond To Micron Stock As AI Memory Demand Surges
- The AI Money Machine: 55% of Investment in AI Companies Comes From Other AI Companies
- Capcom plans to evolve RE Engine into an "AI-generation game engine"
- What Do Alphabet (GOOGL) AI Safety Pact And Antitrust Wins Mean Now?
Comments
Please log in to post a comment.