Latest / DX Today | No-Hype Podcast & News About AI & DX / AI News Today: Meta Launches Muse Wonder, Its First Frontier Model

Transcript

AI News Today: Meta Launches Muse Wonder, Its First Frontier Model · DX Today | No-Hype Podcast & News About AI & DX

• 0:00
  SPEAKER_05: It's Tuesday, October 13th, 2026.
• 0:03
  SPEAKER_05: You're listening to the DX Today AI Daily Brief.
• 0:07
  SPEAKER_05: Today, Meta releases its first Frontier AI model.
• 0:10
  SPEAKER_05: Microsoft unveils a new custom chip and a multi-gigawatt rollout.
• 0:15
  SPEAKER_05: And character.ai files for bankruptcy with Disney as the first bidder.
• 0:21
  SPEAKER_05: Let's get into it.
• 0:22
  SPEAKER_00: Meta has released its first frontier AI models.
• 0:26
  SPEAKER_00: On Monday, Meta's Superintelligence Lab introduced Muse Wonder and a larger version called MuseWonderful, which the company calls its most capable models to date.
• 0:35
  SPEAKER_00: They are built to power personal and professional agents, handling tasks like navigating an insurance claim, making sense of a lease, or building a working app.
• 0:44
  SPEAKER_00: Meta says MuseWonder now outscores leading models from OpenAI and Anthropic on a memory benchmark.
• 0:50
  SPEAKER_00: Though those results come from the company's own testing, Meta's AI chief Alexander Wang said the model can carry information across days, weeks, and months, so users do not have to repeat themselves.
• 1:02
  SPEAKER_00: MuseWonder is available now to developers through the Meta model API and Muse code, and will soon power the Muse app, starting with subscribers.
• 1:11
  SPEAKER_02: Next, a chip story.
• 1:13
  SPEAKER_02: Microsoft is betting bigger on its own silicon.
• 1:16
  SPEAKER_02: The company announced Maya 300, the third generation of its custom AI chip, and said it will expand Azure Maya systems to multi-gigawatt deployments beginning in 2027.
• 1:31
  SPEAKER_02: Maya 300 is designed for inference first, tuned for long context, memory, heavy, and agentic workloads.
• 1:41
  SPEAKER_02: Microsoft says it should deliver 40% more tokens per megawatt than leading graphics processor systems and up to seven times the rack level throughput of the current Maya 200.
• 1:54
  SPEAKER_02: Each chip carries eight stacks of high bandwidth memory.
• 1:58
  SPEAKER_02: With the rollout, Maya becomes the primary inference system for Microsoft products such as Copilot.
• 2:05
  SPEAKER_02: Microsoft is not walking away from Nvidia, but the move reduces how much it depends on outside chips.
• 2:12
  SPEAKER_04: Now a chatbot pioneer stumbles.
• 2:15
  SPEAKER_04: Character.ai has filed for bankruptcy.
• 2:18
  SPEAKER_04: The chatbot company said Monday it started a voluntary Chapter 11 case in Delaware to run a court supervised sale of substantially all of its assets.
• 2:27
  SPEAKER_04: Disney has signed on as the stalking horse bidder, offering$20 million for a non-exclusive perpetual license to certain character.ai, intellectual property, and related legal claims.
• 2:39
  SPEAKER_04: That bid sets a floor, and rival buyers can still top it.
• 2:43
  SPEAKER_04: The filing follows mounting legal pressure over teen safety and a copyright dispute in which Disney had demanded the company stop chatbots that mimicked its characters.
• 2:53
  SPEAKER_04: Character.ai says its app and website will keep running during the process, and it expects the sale to close in late November or early December.
• 3:02
  SPEAKER_04: Turning to model security.
• 4:15
  SPEAKER_03: Nvidia and Cisco also unveiled a new pluggable optical module design built for 400 gigabits per lane.
• 4:23
  SPEAKER_03: They say it delivers four times the bandwidth density and about 30% better power efficiency than today's 200 gigabit optics.
• 4:30
  SPEAKER_03: Nvidia added that Meta is adopting its Spectrum 6 Ethernet switches, and that Google plans to support its own Falcon transport on future NVIDIA networking hardware.
• 4:40
  SPEAKER_03: Some analysts note that NVIDIA is opening the plumbing while keeping its most profitable chips firmly in-house.
• 4:46
  SPEAKER_05: Staying with that summit.
• 4:48
  SPEAKER_00: AMD used the same summit to chase the agent boom.
• 4:51
  SPEAKER_00: The chipmaker introduced Helion, an open rack design built for agentic AI, where software agents plan and carry out multi-step tasks.
• 5:00
  SPEAKER_00: Helion comes in two versions.
• 5:02
  SPEAKER_00: A processor-only version for hyperscale clouds packs up to 32,768 cores in a single rack, using sixth generation APIC chips.
• 5:13
  SPEAKER_00: An all-in-one enterprise version mixes those processors with instinct graphics processors so a company can run agents and model inference in one rack.
• 5:23
  SPEAKER_00: AMD also built in a token router that directs agents model calls across frontier and open source models.
• 5:30
  SPEAKER_00: Cisco, Dell, Foxconn, Hewlett-Packard Enterprise, Lenovo, Supermicro, and others are developing systems based on the design, which AMD says ships next year.
• 5:43
  SPEAKER_02: Over to Enterprise Services.
• 5:45
  SPEAKER_02: Indian Technology Services Giant, HCL Tech, is going all in with Anthropic.
• 5:52
  SPEAKER_02: On Monday, HCL Tech signed a global premier partnership and said it will set up a dedicated Anthropic business unit and center of excellence to speed enterprise deployments of Clawed, with a focus on safety, cybersecurity, and governance.
• 6:10
  SPEAKER_02: Its software arm, or HCL Software, will embed Claude and Clawed code into selected enterprise products.
• 6:19
  SPEAKER_02: The announcement landed alongside quarterly results.
• 6:22
  SPEAKER_02: HCL Tech reported net income up about 21% from a year earlier in the second quarter of its fiscal year, with record new deal bookings of$3.3 billion.
• 6:36
  SPEAKER_02: The partnership reflects a broader race among Indian outsourcing firms to rebuild their services around AI agents as clients push for automation over headcount.
• 6:49
  SPEAKER_04: Next, a funding story.
• 6:50
  SPEAKER_04: Also in India, Sarvom AI may be about to double its value in four months.
• 6:56
  SPEAKER_04: The Economic Times reports the startup is in early talks to raise$250 to$300 million at a valuation of about$3 billion.
• 7:06
  SPEAKER_04: In June, Sarvum raised more than$300 million at a$1.5 billion valuation.
• 7:13
  SPEAKER_04: If the new round closes on those terms, Sarvum would become India's most valuable AI startup.
• 7:20
  SPEAKER_04: The company builds Indian language, AI models, and applications for government and enterprise customers.
• 7:26
  SPEAKER_04: And it recently hired the former head of Google Cloud in India as its president.
• 7:31
  SPEAKER_04: People familiar with the talks say Sarvum wants to keep a significant share of ownership with Indian investors.
• 7:38
  SPEAKER_04: The amount and the valuation could still change.
• 7:41
  SPEAKER_04: Now, the cost of compute.
• 7:43
  SPEAKER_01: AI costs are now hitting paychecks at Zoho.
• 7:46
  SPEAKER_01: Founder Sridhar Vembu said the Indian software company will skip its usual annual salary increases this year.
• 7:53
  SPEAKER_01: In a post on X, Vembu blamed sharply rising AI costs, a fourfold jump in server prices, and an uncertain geopolitical backdrop.
• 8:03
  SPEAKER_01: He said Zoho has already cut its marketing budgets, has not grown its headcount in 18 months, and is working hard to avoid mass layoffs.
• 8:12
  SPEAKER_01: Vembu also warned that the outlook for the software business is uncertain.
• 8:17
  SPEAKER_01: The move is a notable signal from a company long known for steady, self-funded growth.
• 8:23
  SPEAKER_01: It shows how the rising price of computing power and AI infrastructure is starting to reshape budgets and paychecks well beyond the big model makers.
• 8:33
  SPEAKER_05: Next, a new investor.
• 8:35
  SPEAKER_03: Chinese AI lab DeepSeek is putting money into robotics.
• 8:39
  SPEAKER_03: Bloomberg reports DeepSeek has agreed to lead a seed round worth tens of millions of dollars for Winverse AI, a month's old startup building AI for robots.
• 8:49
  SPEAKER_03: Winverse was founded this year by Hu Han, a former top 10 cent scientist who led its vision language and embodied foundation model teams and who previously worked at Microsoft Research Asia and Baidu.
• 9:02
  SPEAKER_03: Other backers, including Maytoon's Long Z investments, would value the startup at about$400 million after the investment, according to people familiar with the matter.
• 9:13
  SPEAKER_03: The deal is not final, and the terms could change.
• 9:16
  SPEAKER_03: It would be one of Deep Seek's largest startup bets, funded from its own raise of more than$12 billion.
• 9:24
  SPEAKER_05: To the factory floor.
• 9:25
  SPEAKER_00: Japan's OMRON is buying its way into physical AI.
• 9:28
  SPEAKER_00: The industrial automation company agreed to acquire American industrial computer maker on Logic, along with its parent company, InnerLogic, for an enterprise value of$675 million.
• 9:40
  SPEAKER_00: On, Logic builds rugged computers that sit on the factory floor, handling heat, dust, and vibration while processing sensor data in real time instead of waiting on the cloud.
• 9:51
  SPEAKER_00: Omaron plans to pair that edge hardware with its own precision control technology to build physical AI platforms and to grow its customer base in the United States.
• 10:01
  SPEAKER_00: More than 90% of OnLogic's sales come from the Americas and Europe.
• 10:05
  SPEAKER_00: OnLogic will keep operating independently with its current management and brand.
• 10:09
  SPEAKER_00: The deal still needs regulatory approval.
• 10:12
  SPEAKER_05: Finally, a policy story.
• 10:14
  SPEAKER_02: And in Britain, the government is stepping up its defenses against AI risks.
• 10:20
  SPEAKER_02: AI Minister Kanishka Narayan said on Monday that the country's AI Security Institute has access to the latest advanced models, including Google's newest Gemini model, and is running rigorous safety evaluations with the United States government and major developers.
• 10:39
  SPEAKER_02: The government will expand its use of Frontier AI to harden its own cyber defenses, building on two pilots that found and fixed more than a thousand vulnerabilities in government systems.
• 10:54
  SPEAKER_02: It will also create a specialist team to respond to major incidents involving agentic AI Narran warned that it is likely the country's cyber defenses will be breached.
• 11:08
  SPEAKER_02: And he told AI companies, if you cannot adequately control the risks, do not release the system.