The Morning Download: Scaled Cognition Tackles AI Reliability

Good morning. The signs of AI’s influence are omnipresent this morning, from inflation trends to workplace shifts and memory chip maker Micron's seizure of stock market leadership.

All of those broader shifts rest on a powerful surge of innovation in AI that keeps moving ahead. This effort extends well beyond the development of increasingly powerful frontier models by the major AI labs, and into an ecosystem of startups that are working on new approaches to model architecture of the infrastructure around models.

Dan Roth, co-founder and CEO of one such AI lab, Scaled Cognition, told me that the startup has raised $100 million in a Series A round led by Khosla Ventures. Genesys, a provider of cloud-based AI customer experience technology, also invested in the round, which valued the company at $750 million.

Scaled Cognition’s mission is to make AI more reliable, so that it can be fully trusted to handle tasks in areas of business where there’s no margin for error. Those are the kinds of efforts that keep driving AI’s influence more deeply into the economy. See below for more details -- and you can read our full exclusive story here.

Scaled Cognition, based in Mountain View, Calif., was founded by Roth and Chief Technology Officer Dan Klein, a natural language processing researcher and professor of AI at the University of California, Berkeley. They sold their prior startup, Semantic Machines, to Microsoft in 2018.

“These frontier models that are out there are amazing—they’re intelligent in so many different ways—but they’re sort of like schizophrenic geniuses,” Roth told me. “They can create incredible answers, and then you can ask them the same question a second time and get a completely different answer that … might not even be correct.

“We really believe that for these systems to really be useful, you have to be able to trust them. And in order for you to trust them, they have to be provably reliable,” he said.

Roth and Klein set out to design an alternative AI architecture that delivers reliably correct results. The result was APT, or Agentic Pretrained Transformer, as their flagship model is called.

Beyond the model, Scaled Cognition has also built a platform for enterprise AI deployment that includes agentic tooling, live agent monitoring and simulation and evaluation frameworks, the company said.

It is targeting customer experience as a first market, but Genesys is already using APT within its Genesys Cloud platform for agentic virtual agent capabilities.

Roth said large language models rely on token-by-token text prediction, optimizing for linguistic plausibility. The model is detached from external reality and lacks an inherent understanding of whether the output is correct. Scaled Cognition addresses AI hallucination with a model that predicts structured objects, such as programs and system queries, in addition to token streams.

Scaled Cognition’s architecture directs different portions of a query to the most appropriate system, depending upon variables such as the need for extreme reliability, according to venture capital investor Vinod Khosla, founding partner of Khosla Ventures.

“It is a separate model for those parts of the system that need real reliability and cannot be subject to hallucination,” he said.

People get ready. We're far enough into the current cycle that business leaders are suddenly confronting the realization that perhaps they aren't as ready to scale AI as they thought. 

IT infrastructure services provider Kyndryl surveyed some 1,100 senior business leaders across eight markets for their take on the state of AI in their organizations. Their responses, captured in the 2026 People Readiness Report released today, reveal a gap between the perceived readiness of their tech capabilities and that of their people.

Just 23% of organizations think their workforces are fully ready for AI, a six-point drop from last year, according to the report. Additionally 79% agree that the speed of AI will outpace their organizations’ workforce.

The gap between employee readiness and the technology was an "aha moment" for Kyndryl's CIO, Kim Basile, who shared her take on the report with the WSJ Leadership Institute.

"I've said this from day one as we talk about AI, it is 100% about the people. It is about the people's ability to use this technology," she said. 

"The wider the gap, the more the lack of trust that's there... and the harder it is going to be for organizations to scale," she said. 

The study also spotlighted a small group, called "pacesetters," addressing that challenge, not through more technology but in how they prepared their people.

"The pacesetters are definitely redesigning the roles around AI that align with the business process workflow and really evaluating that," said Basile. "They've got strong change management about what they're asking to do. So it's all centered around the people. And then there's a look that says, governance and trust go hand in hand with that."

Basile said that approach upends the usual playbook for adopting new technology.

“If you think back, most technology, there was your set of your engineers and your developers, and then there was everyone else," she told the WSJLI.

“What AI has done is broken that up, and everyone at some point can use AI. So in order to let them do that, you want them to really be able to understand what those guardrails are. And where those rules of the road come into play. And you see that also come out in the Readiness Report about how guardrails and trust are key.”

• America's massive AI build-out is emerging as a new driver of inflation, pushing up prices on everything from smartphones and videogame consoles to electricity, the WSJ reports. While many economists believe AI could eventually lower inflation by boosting productivity, that effect is likely years away. In the meantime, 81% of economists in a recent survey expect the build-out to add to inflation over the next year.

• Shares in Micron Technology are soaring 17% premarket trading, after the memory chip-maker Wednesday beat expectations in virtually every closely watched financial metric, the WSJ reports. Micron said it expects the memory-chip shortage to go beyond 2027. The blockbuster earnings arrived after a slump in tech shares that rippled from South Korean chip makers to American fiber-optic cable providers, the WSJ reports.

• Anthropic has accused Alibaba of conducting "the largest known distillation attack" to date, alleging in a letter to U.S. Senators Tim Scott and Elizabeth Warren that the Chinese company used nearly 25,000 fake accounts and 29 million exchanges to illicitly extract capabilities from its Claude AI model, WSJ reports.

• The European Union said Amazon and Microsoft should be treated as gatekeepers for their cloud computing services under the bloc’s Digital Markets Act, a designation that requires large tech companies to take extra steps to ensure they are not stifling competition, WSJ reports.

• A bipartisan coalition called RAISE US backed by major employers like Amazon, Microsoft, and Bank of America launched Thursday to prepare the U.S. workforce for AI-driven job disruption, WSJ reports. Beyond retraining programs, the group aims to rethink longstanding policies like unemployment insurance and test ways to help displaced workers transition into new roles and fields.

WATCH | How will.i.am became a professor at Arizona State University. The tech entrepreneur and recording artist explains how his AI company led to an invitation from Arizona State University President Michael Crow to integrate the tech into the classroom and teach at the university. 

Panic and Popcorn. “This is an emergency,” Rory Blundell, CEO of AI security firm Gravitee, told me, while a gathering of suited Manhattan CTOs sipped champagne and ate tiny cheese sandwiches a few feet away.

We were in an underground cocktail cave at the Bryant Park Hotel where Gravitee was hosting the screening of a 25-minute documentary it commissioned, “AI Agents: How Do We Take Back Control?"

Companies rolling out autonomous agents without a full understanding of the security implications are setting themselves up for disaster, Blundell said, adding that he knew of one company where an agent deleted the corporate calendar. Not quite apocalyptic, but as technologies like robotics and quantum computing mature and converge with AI, the stakes for rogue agents are only to get bigger, he said.

After the reception, which included a red carpet, executives moved into a screening ro