Technology

➚ MOVE FAST: Fission. Constellation Energy inked its second major nuclear power deal with a Big Tech company, supplying Meta for 20 years. AI’s thirst for power is fueling the boom, sending Constellation’s stock up about 30% this year.

➘ BREAK THINGS: Cohere. The AI startup is looking to raise $500 million in a bid to catch up to bigger rivals OpenAI and Anthropic. That would still give it just a $6.5 billion valuation, compared to $300 billion for Sam Altman’s firm.

How vibe coding is tipping Silicon Valley’s scales of power

In almost every way, Leo Paz is following a well-worn path to Silicon Valley success. The 27-year-old Canadian earned a degree in software engineering, co-founded a startup that was accepted into the vaunted Y Combinator accelerator program, and moved to San Francisco to build it.

But in another way, he’s unlike any of his predecessors: Since landing in the world capital of software development, he’s let AI do all the coding. Paz has been spending 14-hour days instructing large language models to do the work for him, “vibe coding” his sales agent startup Outlit all the way to Y Combinator demo day in April.

“I don’t think YC knew exactly how much code was being written by LLMs,” he told Semafor. It found out earlier this year after a founder survey. “They were kinda shocked,” he added.

In March, Y Combinator CEO Garry Tan said 25% of startups in that winter’s class generated 95% of their code using AI tools. The numbers are likely to be much higher in the next batch, which culminates with a demo day this month.

Outside of YC, Pukar Hamal, CEO of SecurityPal — which helps companies automate responses to security-related questions during the sales process — is vibe coding agents to fill some open positions rather than hiring personnel.

SecurityPal needed a product manager, and instead of creating a job posting, Hamal spent three hours vibe coding an agent named Penelope, depicted as a brunette mountaineer wearing aviator goggles and a fur-lined bomber jacket (a nod to the company’s office in Nepal). “Penelope is going to be a very integral part of how I as a CEO make decisions,” he told Semafor. “She’s also going to be part of my leadership meetings, where we can interrogate her on different ideas we have.”

The long, acrimonious fight between payroll software companies Deel and Rippling has taken another turn. On Tuesday, Deel accused its rival in a lawsuit of corporate espionage by directing an employee to pose as a Deel customer and collect information on the company — including pricing data, benefits policies, and templates for hiring and paying international workers. Over six months last year, the Rippling worker passed information onto his employer, which then created copycat products, Deel’s filing said.

“Rippling is unwavering in our commitment to fair competition and the highest ethical standards,” a Rippling spokesperson said in a statement to Semafor. “Notably, in this new complaint Deel backtracks from some of its original bogus claims, showing once again that with Deel every accusation is just an admission of its own misconduct.”

Similar spying claims came from Rippling earlier this year, culminating in an employee flushing his phone down the toilet. According to a Rippling lawsuit, an employee at the company was being paid to spy for Deel, sharing Rippling’s internal information like pricing, training materials, and conversations with customers. After presenting the employee with a court order for his phone in March, he sent his phone down the bathroom pipes. The employee, who is no longer with the company, has since admitted to spying and is cooperating with Rippling’s lawyers. Deel filed to dismiss Rippling’s corporate espionage case.

In a comment to Semafor, a Deel spokesperson said, “Rippling is relying on a witness who has provided testimony pursuant to a cooperation agreement that Rippling refuses to disclose.” Deel’s new filing, however, is “based on incontrovertible evidence” of Rippling stealing proprietary documents, they said.

The flood of accusations and pointed fingers underscores the long-running antagonism between the companies, which offer similar services and compete for customers. It appears to be coming to a head with the recent lawsuits, which are airing out years of dirty laundry.

Deel’s lawsuit in part relies on Rippling CEO Parker Conrad’s reputation. At his previous startup Zenefits, Conrad developed a “company culture of pressuring and bullying employees to cut corners and do the wrong thing,” his replacement CEO David Sacks wrote in a memo at the time. Conrad posted his response to the personal attacks made in Deel’s original filing in April: “They discovered I was CEO of this co called Zenefits and guys, it.did.not.end.well!”

Whether the spectacle impacts customer retention is still to be determined.

— Rachyl Jones

Anthropic is going toe-to-toe with the Trump administration on artificial intelligence policy, according to a person familiar with the matter, irking White House officials and bucking the trend of AI companies seeking closer ties with the government.

The $60 billion Silicon Valley company responsible for some of the world’s most advanced AI models has been lobbying members of Congress to vote against a federal bill that would preempt states from regulating AI, according to two people familiar with the matter. The bill is a key component of the Trump administration’s efforts to clear the way for US companies to advance the technology.

One of Anthropic’s advisors also pushed against the recent AI deal with Gulf states, which would see US technology flow to the region in exchange for an influx of investment money flowing the other direction.

The efforts have angered staffers within the Trump administration working on AI policy who have come to see Anthropic as an obstacle. At a recent meeting at the White House, according to one person familiar with the matter, officials lamented Anthropic’s efforts and called out the company’s hiring of several former Biden administration staffers, including Elizabeth Kelly and Tarun Chhabra. Ben Buchanan, another Biden official, is an adviser.

Anthropic also employs policy staff who worked for Republicans, like legislative analyst Benjamin Merkel and lobbyist Mary Croghan.

Anthropic declined to comment.

As electricity demand soars — driven by the rapid expansion of data centers and AI — pressure is mounting to scale secure and reliable energy resources.

Join Semafor for a timely conversation with Chairman Mark Christie, FERC; Rep. Brett Guthrie, R-Ky.; and Aamir Paul, President of North American Operations at Schneider Electric, as they discuss how the new Administration plans to accelerate domestic energy production — and whether current infrastructure is up to the task. The discussion will also explore the innovative policies and technologies that could help close the growing supply-demand gap.

June 11, 2025 | Washington, DC | RSVP

The Wall Street Journal’s scoop that Meta plans to offer personalized AI-generated ads by the end of the year points to a brave new world for advertising — and for news.

Increasingly sophisticated AI tools mean that companies — in this case, advertisers or Meta — are close to being able to generate, in real time, sales pitches specifically tailored to individuals. Per the Journal: “A person seeing an advertisement for a car in a snowy place, for example, might see the car driving up a mountain, whereas a person seeing an ad for that same car in an urban area would see it driving on a city street.”

It doesn’t have to stop there. People who have a history of buying red cars might see an ad featuring one; if they like classical music, maybe that’s what the soundtrack plays; and so on. We’re on the brink of being able to create entirely personalized messages, limited only by compute power and how much data exists about the end user and their preferences.

That promises to upend the advertising industry. But what will it mean for news? How soon can we create much more personalized versions of an article? Perhaps you’d get a shorter, audio version when your phone detects you’re driving or running; or one that omits background information it knows you read yesterday; or one that weaves in more context when the AI believes it’s a topic you’re unfamiliar with. Or a bot could rewrite the story using language at a level you’re comfortable with, or offer a tool that allows you to engage in an extended conversation to find out more.

But there are also real questions about whether a world where everyone is in their own personal news bubble will increase polarization further, even if all the information they receive is entirely accurate. It’ll doubtless be a tool, too, for anyone who wants to personalize propaganda — or create precisely tailored political messaging that will make today’s micro-targeting look like a crude sledgehammer.

A researcher at Virginia’s George Mason University is developing an augmented reality firefighter training program that simulates real-world scenarios without the real-world dangers, the university reported. Last year, 17% of firefighter deaths occurred during training exercises, which includes suppressing live fires and drills in climbing ladders.

The program creates digital twins of existing rooms, using AI to simulate a realistic outbreak of fire and place civilians where they are most likely to be. The scenarios vary in difficulty to capture all levels of firefighter experience.

“I don’t ever see us replacing real fire training, but this is a means to augment the training to improve the health and safety of the firefighters, as well as make training easier to deploy in a variety of buildings and circumstances,” the researcher, Craig Yu, told the university. “AI allows us to rapidly analyze human performance and then adaptively synthesize training programs to address weakness we observe.”

US President Donald Trump laid out a new US approach to the Middle East during his Gulf visit last month, abandoning lectures in favor of engagement with partners who share his vision for security, stability, and prosperity, Jason D. Greenblatt, Trump’s former Middle East envoy, wrote in a Semafor column.

Whereas US policy lurched “between overcommitment and neglect, the Trump Doctrine offers something genuinely new: a sustainable framework for American engagement based on shared prosperity rather than shared enemies,” Greenblatt wrote.