Good morning. I’m in light rain in Las Vegas, attending Workday Rising—the annual customer conference for this human resources software company.
To honor last week’s launch of the brand-new Fortune Future 120 list, I moderated a dynamic discussion on how companies can drive growth from within—in the so-called Age of Intelligence, where technology, talent, and, of course, methodology are all critical. What insights did Workday’s CIO Rani Johnson, Samsara’s CIO Stephen Franchetti, and BCG Managing Director and Partner Ketil Gjerstad share with top IT executives on embracing AI within enterprises?
As usual, let's start with a haiku:
Let everything be safely said: let a hundred flowers bloom, then prune most of them back.
Here is today's tech news. — Andrew Nusca
Headline Story
Meta launches Muse Wonder in response to cutting-edge AI for agents.
For Meta, the parent company of Facebook that has been constantly shifting direction, two unresolved questions have surrounded its volatile AI business.
First: Will the restructuring of the Superintelligence Labs team in 2025 lead to more competitive AI outputs?
Second: Will the company's $145 billion capital expenditure spree yield a return?
This year, Meta has spent most of its time addressing the first question. The company first captured everyone’s attention in April with the launch of its Muse Spark series of large language models, followed by the introduction of its personal AI agent, Muse, last month.
On Monday, the company further expanded its offerings by launching Muse Wonder and Muse Wonder Full, describing them as cutting-edge models designed for agents.
You might ask: Doesn’t Muse Spark also claim to have agent capabilities? Indeed it does—but Meta’s new Muse Wonder dual model is larger and more intelligent, and therefore slower than the former.
This is precisely what sets cutting-edge AI models apart, which is why Muse Wonder was trained on Meta’s megawatt-scale Prometheus data center. (And yes, it will power certain experiences within Muse agents and also be used to develop future specialized AI models.)
In a detailed blog post listing multiple benchmark tests, Meta stated that its new model can compete with leading competitors such as OpenAI’s GPT-6 Astra, Anthropic’s Claude Fable 5.1, and Opus 5.5.
The company also highlighted two new capabilities of these models: Active Memory, which Meta describes as the ability to retain context across conversations; and Active Focus, which Meta describes as resilience when the initial approach fails to complete a task.
As for the thorny second issue—return on investment? Unlike its peers, Meta’s AI efforts are primarily focused within its consumer-facing platform ecosystem. While its agents are unlikely to handle your department’s expense reports or revise corporate contract terms—your employer will likely prefer enterprise-grade alternatives—they can certainly help you file insurance claims, understand apartment leases, or build a mobile app for your small business.
Just last week, the article noted that Meta’s personal AI agents could contribute up to 8% of Meta’s revenue by 2030, equivalent to approximately $36 billion in sales. Would that be enough to offset the salaries of those AI engineers? Not even close. But no one understands better than Meta that maintaining the status quo isn’t an option. —Andrew Nusca
Mystery Box
We do believe this is an opportunity that will last for a considerable period of time.
Click here to read the full article.
Briefing
- Microsoft has unveiled its custom-built chip: the third-generation AI chip, Maia 300, built on a 2-nanometer process, designed to serve as the company’s primary AI inference system.
- A cyberattack caused a French supermarket's refrigeration system to rise to 50 degrees Celsius within hours.
- Google's new Fitbit Edge fitness tracker: replaces the Charge 6, shipping on October 27 for $180.
- The outgoing head of Microsoft’s science division plans to join... the Mercedes F1 team?
- Epic has delayed the release of an upcoming Disney game and reportedly asked the developers to complete it using AI.
- Since July, Elon Musk has donated over $276 million to Republican-aligned interests.
- Corporate cryptocurrency purchase: "Spread the word that we will support it."
Final glance
What is the biggest obstacle to AI?
According to Hock Tan, CEO of Broadcom, and Lip-Bu Tan, CEO of Intel, who spoke yesterday in Singapore (no relation), the issue is not memory chip supply, not evolving regulations, and not recent failures by autonomous agents in various settings—it’s real money.
Broadcom's CEO said, "The biggest constraint is capital."
Intel's CEO agreed: "If there's a challenge ahead, it's capital and the cost of capital, and in the longer term, people want to see a return on investment."