Meta launched Muse Wonder, its most advanced AI model, alongside a variant called Muse Wonder Full. The company says both models match competitors like Anthropic in reasoning, writing, and coding. Muse Wonder is an AI agent that can act on behalf of users, and it reached the top of Apple's App Store. Access requires a monthly subscription, and the models were delayed after Meta paused development in July.
Meta released Muse Wonder on the Meta Model API and Muse Code, while Muse Wonder Full is limited to paid Muse Code plans. In benchmarks, Muse Wonder Full leads in 6 of 14 rows in Meta's table and scores 85.6 on MemoryBench. Muse Wonder Full costs $50 per month through Muse Code, and the API is OpenAI-compatible. Pivotal Research raised its Meta price target from $790 to $930, citing early adoption gains.
Broadcom CEO Hock Tan warned that securing funding for AI infrastructure could become harder than getting chips. He noted wafer and memory supply is improving, but building advanced AI systems is expensive. Broadcom is in talks to arrange over $50 billion in debt financing to help OpenAI buy custom AI chips and is lending up to $42 billion to Anthropic. Reports also mention a possible $60 billion financing effort for AI chips.
Elon Musk responded to Tan's warning that AI investors are gambling. Tan spoke at a conference in Singapore and said many investors do not understand what they are investing in. Broadcom reported AI chip revenue of $16.7 billion in the third quarter, a 221% increase. A separate Intel CEO also warned about AI infrastructure strain. Research projects US AI spending could reach $10.3 trillion by 2032.
SoftBank issued $11 billion in junk bonds with yields as high as 9.75%, the highest ever for the company. CEO Masayoshi Son is making an all-in bet on AI, including backing OpenAI with up to $64.6 billion. The company chose to borrow rather than sell assets like Nvidia or T-Mobile US. SoftBank shares have fallen 38% since June, and analysts are watching whether AI investments will pay off.
Researchers created LifeSearch, a benchmark with over 200 real-world tasks to test deep research AI agents. The tasks span seven life and productivity areas and include more than 4,500 hours of human annotation. Even the best AI agent, GPT-5.6 Sol, scored only 63.4% on the tests. The benchmark aims to better measure how well AI agents perform practical, everyday research tasks.
Growing use of AI agents is creating strong demand for hardware, especially CPUs. This was a major topic at the Open Compute Project Global Summit in San Jose. Industry leaders highlighted that agentic AI workloads require significant computing power. The trend signals a shift in data center hardware needs as AI agent adoption increases rapidly.
Unapproved AI tools, called Shadow AI, are increasing data breach risks in trucking and logistics. Cybersecurity incidents linked to Shadow AI rose to 43% in 2026, up from 20% the year before, with average breach costs reaching $5.39 million. The NMFTA Cybersecurity AI Governance Framework offers 50 practical controls to help carriers manage AI risk. Experts recommend fleets assign ownership, audit tools, and create clear AI usage policies.
Policymakers are creating new rules to ensure the US dollar remains central as AI agents begin handling financial transactions. These digital agents can spend money around the clock without human approval. Current regulations favor dollar-backed stablecoins, held by major issuers like Circle, Coinbase, and Binance. The US government plans to finalize rules by January 18, 2027, to supervise these companies and prevent financial risks.
Palantir Technologies stock is priced at $208.67, higher than future cash flows suggest. The company has grown its stock value about 11 times over three years. New projects in sovereign AI, such as the Regna AI venture, explain investor willingness to pay this price. However, the stock faces a bear case argument that it is overvalued by 22%, trading at over 56 times sales. Some analysts believe it is undervalued by 42% as it positions itself as a key AI tool for governments.
Mibro launched the AeroPace smartwatch, featuring AI-powered running coaches for beginners. The device offers more than 20 training programs lasting four to 20 weeks. It adjusts plans based on user data like sleep quality, fatigue, and past exercise history. Runners can choose from five virtual coaching personalities and talk to them using the Just Talk feature. After every run, the system analyzes performance and recovery to suggest improvements.
Brazilian startup Imersa raised $49,700 in seed funding to build an AI sales copilot. The company plans to use the money to expand its team and improve product technology. The AI tool helps sales teams automate routine tasks so they can focus on high-value activities. Imersa has gained market traction with several clients. The founders have strong backgrounds in AI and sales.
Key Takeaways
- Meta launched Muse Wonder and Muse Wonder Full, with the latter costing $50 per month and scoring 85.6 on MemoryBench.
- Broadcom is arranging over $50 billion in debt financing for OpenAI and lending up to $42 billion to Anthropic.
- Broadcom reported AI chip revenue of $16.7 billion in Q3, a 221% increase.
- SoftBank issued $11 billion in junk bonds with yields up to 9.75% to fund AI bets including OpenAI.
- LifeSearch benchmark found even the best AI agent, GPT-5.6 Sol, scored only 63.4% on real-world tasks.
- AI agents are driving strong CPU demand, shifting data center hardware needs.
- Shadow AI cybersecurity incidents in trucking rose to 43% in 2026, with average breach costs at $5.39 million.
- US plans to finalize AI financial transaction rules by January 18, 2027, focusing on dollar-backed stablecoins.
- Palantir stock trades at over 56 times sales despite 11x growth over three years.
- Imersa raised $49,700 in seed funding for its AI sales copilot tool.
Meta launches Muse Wonder, its most advanced AI model yet
Meta announced two new AI models called Muse Wonder and Muse Wonder Full on Monday. The company says these models match competitors like Anthropic in reasoning, writing and coding. Muse Wonder is an AI agent that can act on behalf of users, and it reached the top of Apple's App Store. Access to Muse Wonder requires a monthly subscription, and the models were delayed after Meta paused development in July.
Meta's Muse Wonder highlights a major AI trade contradiction
Meta Platforms says its Muse Wonder model strengthens its position in frontier AI development. The company believes its consumer-focused strategy and wide distribution support long-term growth. However, rising AI infrastructure costs and economic pressures could hurt Meta's free cash flow and lead to a lower stock valuation after Q3 earnings.
Meta price target raised to $930 as AI investment gains look promising
Pivotal Research raised its Meta price target from $790 to $930 and kept a Buy rating. Analyst Jeffrey Wlodarczak says early adoption of the Muse assistant and Muse Wonder model shows Meta's AI investment returns are appearing. Muse reached 2.8 million installs in 12 days, but its day-14 retention rate was lower than ChatGPT's. Meta reports Q3 earnings on October 21, which will show Muse's first real revenue impact.
Meta releases Muse Wonder and Muse Wonder Full with pricing details
Meta released Muse Wonder and Muse Wonder Full, its most capable AI models so far. Muse Wonder is available on the Meta Model API and Muse Code, while Muse Wonder Full is limited to paid Muse Code plans. In benchmarks, Muse Wonder Full leads in 6 of 14 rows in Meta's table and scores 85.6 on MemoryBench. Muse Wonder Full costs $50 per month through Muse Code, and the API is OpenAI-compatible.
Broadcom CEO warns AI funding may be harder than chip supply
Broadcom CEO Hock Tan said securing funding for AI infrastructure could become a bigger challenge than getting chips. He noted that wafer and memory supply is improving but building advanced AI systems costs a lot. Broadcom is in talks to arrange over $50 billion in debt financing to help OpenAI buy custom AI chips and is lending up to $42 billion to Anthropic. Reports also mention a possible $60 billion financing effort for AI chips, and Anthropic could become Broadcom's largest chip customer in 2027.
Musk Counters Broadcom CEO's AI Casino Warning
Elon Musk responded to Broadcom CEO Hock Tan's warning that AI investors are gambling. Tan spoke at a conference in Singapore and said many investors do not understand what they are investing in. Broadcom reported AI chip revenue of $16.7 billion in the third quarter, a 221% increase. A separate Intel CEO also warned about AI infrastructure strain. Research projects US AI spending could reach $10.3 trillion by 2032.
SoftBank Sells Record Junk Bonds to Fund AI Bets
SoftBank issued $11 billion in junk bonds with yields as high as 9.75%, the highest ever for the company. CEO Masayoshi Son is making an all-in bet on AI, including backing OpenAI with up to $64.6 billion. The company chose to borrow rather than sell assets like Nvidia or T-Mobile US. SoftBank shares have fallen 38% since June, and analysts are closely watching whether AI investments will pay off soon.
LifeSearch Tests AI Research Agents on Real Tasks
Researchers created LifeSearch, a benchmark with over 200 real-world tasks to test deep research AI agents. The tasks span seven life and productivity areas and include more than 4,500 hours of human annotation. Even the best AI agent, GPT-5.6 Sol, scored only 63.4% on the tests. The benchmark aims to better measure how well AI agents perform practical, everyday research tasks.
AI Agents Drive Huge Hardware Demand, Experts Say
Growing use of AI agents is creating strong demand for hardware, especially CPUs. This was a major topic at the Open Compute Project Global Summit in San Jose. Industry leaders highlighted that agentic AI workloads require significant computing power. The trend signals a shift in data center hardware needs as AI agent adoption increases rapidly.
Trucking Industry Faces AI Governance and Shadow AI Risks
Unapproved AI tools, called Shadow AI, are increasing data breach risks in trucking and logistics. Cybersecurity incidents linked to Shadow AI rose to 43% in 2026, up from 20% the year before, with average breach costs reaching $5.39 million. The NMFTA Cybersecurity AI Governance Framework offers 50 practical controls to help carriers manage AI risk. Experts recommend fleets assign ownership, audit tools, and create clear AI usage policies.
US Rules Aim to Secure Dollar in AI Economy
Policymakers are creating new rules to ensure the US dollar remains central as AI agents begin handling financial transactions. These digital agents can spend money around the clock without human approval, which changes how payments and money work. Current regulations favor dollar-backed stablecoins, which are already held by major issuers like Circle, Coinbase, and Binance. The US government plans to finalize rules by January 18, 2027, to supervise these companies and prevent financial risks. Officials also want to coordinate with international partners to make sure these payment systems work across different countries.
Palantir Stock Trades High Despite AI Growth Plans
Palantir Technologies stock is currently priced at $208.67, which is higher than what its future cash flows suggest it should be worth. The company has grown its stock value by about 11 times over the last three years, creating a large premium for investors. New projects in sovereign AI, such as the Regna AI venture, explain why investors are willing to pay this high price. However, the stock faces a bear case argument that it is overvalued by 22% because it trades at over 56 times its sales. Conversely, some analysts believe the stock is undervalued by 42% as it positions itself as a key AI tool for governments.
Mibro Watch Adds AI Running Coaches for Beginners
Mibro has launched the AeroPace smartwatch, which features AI-powered running coaches designed for beginner runners. The device offers more than 20 training programs that last between four and 20 weeks. It adjusts plans based on user data like sleep quality, fatigue levels, and past exercise history. Runners can choose from five different virtual coaching personalities and talk to them using the Just Talk feature. After every run, the system analyzes performance and recovery to suggest specific improvements for the next session.
Brazil Startup Imersa Raises Seed Money for AI Sales Tool
Imersa, a Brazilian startup, has raised $49,700 in seed funding to build an AI sales copilot. The company plans to use this money to expand its team and improve its product technology. The AI tool is designed to help sales teams automate routine tasks so they can focus on high-value activities. Imersa has already gained market traction with several clients using its software. The founders have strong backgrounds in AI and sales, and they believe their product will make sales teams more efficient.
Sources
- Meta debuts Muse Wonder, its most advanced AI model yet
- Meta Platforms: Muse Wonder Exposes The AI Trade's Biggest Disconnect
- Meta Price Target Raised to $930 as Analyst Says AI Investment Returns Are Coming into View
- Meta Launches Muse Wonder and Muse Wonder Full: Benchmarks, Pricing and How to Use It
- Broadcom CEO Delivers Stark Warning on Artificial Intelligence
- Elon Musk Fires Back at Broadcom CEO Hock Tan's AI Investment Warning: 'No Crying in the Casino'
- SoftBank junk bond sale raises stakes for CEO Son's all-in AI bet
- LifeSearch: Benchmarking Deep Research Agents for Real-World Life and Productivity
- Agentic AI is Hungry for Hardware
- AI Governance in Trucking: Managing Shadow AI and Fleet Risks
- When AI agents become economic actors, what happens to money and the financial system?
- Palantir (PLTR) Stock May Trade At A Premium Despite Sovereign AI Buzz
- AI-Powered Running Coaches
- Dealroom.co | Brazil's Imersa raises $49,700 seed for AI sales copilot, targets more
Comments
Please log in to post a comment.