• I recently had a procedure that restricted my physical activity for two weeks.
• Afterward, I was disappointed to see a lower number on the scale — and that was new to me.
• I realized that strength training had changed my relationship with my body, and with the scale.

A few weeks ago, I stepped on the scale and saw that my weight had gone down.

My younger self would have looked at that number, and it would have set the tone for a good-mood kind of day. She would have considered it proof of success. I probably wouldn't have admitted it back then, but deep down, I felt that smaller was better. A lower number was better. I may have understood, even then, that weight wasn't the whole picture, but knowing something intellectually and believing it about your own body are very different things.

But, now as an adult, a dietitian, and an avid strength trainer, I looked at that lower number and felt… disappointed.

My reaction to the scale showed me that my perspective had changed

I recently underwent a procedure that restricted physical activity for two weeks, which was honestly the hardest part of the whole experience. When I saw the number on the scale drop, my first thought wasn't, "Great!" It was, "Man…that sucks… I hope I haven't lost muscle."

That was an eye-opener. I realized then just how much strength training had changed my relationship with the scale, and honestly, with my body as a whole.

I knew weight wasn't everything, but I finally understand what that means

I'm a registered dietitian, so I know now, of course, that body weight alone tells us very little about overall body composition. A normal scale can't distinguish between the loss or gain of muscle, fat, water, and other factors.

I also know how important it is to maintain muscle and strength for physical function, metabolic health, and healthy aging. I talk about this stuff all the time, but how it affects your psyche is a different story.

In my 20s, I may have known this on a basic level, but there was a certain disconnect between my understanding and my relationship with the scale.

Now, of course, my knowledge base has grown, but what has really made a massive difference in my relationship with the scale is strength training. I stopped chasing the smallest version of myself and started chasing strength and endurance.

Today, the goals are to exceed PRs in the gym, lift heavier weights, do more pull-ups, and see the muscle I've put so much time, energy, and work into developing. Therefore, instead of losing weight, maintaining or even sometimes gaining weight can be consistent with those goals.

I also think about aesthetics differently now

I'd be lying if I said aesthetics no longer matter to me. Of course they do; everyone wants to look good. All those years of constant messaging about women's bodies, and positive (or negative) feedback, don't just disappear because you learn more about physiology.

But even the way I think about how I look has changed, too. When it comes to aesthetics, I look for different things in my body to improve. I'm much more interested now in muscle definition, shape, and looking strong and healthy than I am in whether a particular bone is visible or whether I've reached some arbitrary number on the scale.

I still weigh myself often. I don't think the scale itself is the enemy; on the contrary, actually. But now, I have a healthy relationship with the scale, something I aim to instill in my clients as well. I just don't assign it the same meaning I once did.

Rather than stress about the number on the scale, I prefer to assess things like: Am I getting stronger? How do I feel? How am I performing? Is my body composition improving? How are my clothes fitting? What do I see when I look in the mirror? This new way of thinking is empowering!

I feel much more confident as well, and that confidence extends far beyond simple body image. When I stepped on the scale after two weeks away from consistently lifting, the scale went down, something my younger self would have celebrated.

But this time, I wanted it back.

• Millennium is offering every employee a personalized AI agent, dubbed a "Digital Twin."
• Twins have their own identities, permissions, and email addresses, the firm said in a blog post.
• Millennium is now working to scale "AI Teammates" as part of its latest technology push.

Millennium employees are getting to know a whole new batch of colleagues.

Izzy Englander's $97 billion hedge fund is offering every employee, including investment staff, a "Digital Twin," a personalized AI assistant who continuously learns from the employee's working style. The firm already has 1,600 AI coworkers invisibly working across Millennium's ranks, and it's now making them available to all 7,000-plus employees in a move to further embed AI into its workforce.

Millennium employees who choose to have a twin can email with them much like they would a colleague, Diana Meditz, the firm's AI product manager, said in a Q&A in a blog post announcing the program. The twin interacts only with its employee, remembering their preferences and handling repetitive tasks, like research, meeting preparation, and inbox management.

Twins aren't clones or extensions of employees, and can't make decisions on their behalf. Rather, they have their own identity, email addresses, permissions, and audit trails, Meditz said in the Q&A. Employees grant their computerized counterparts access on a system-by-system basis, and "guardrails ensure the twin only interacts with its dedicated employee, and its memory, files and workspace live in its own dedicated environment." Meditz said her twin reviews overnight AI news and scans her inbox for relevant internal updates every morning before 6 am.

Vlad Torgovnik, the chief information officer, was the first person at the firm to adopt a digital twin to test how well the concept works. Millennium then scaled up to a pilot group of more than 150 twins; in that first cohort, 97% of employees interacted with their twin each day, the blog post said. Every new hire now gets access to a twin upon joining.

Now that the firm has twins in place, it's working to scale AI teammates, which are built for an entire team. Like the twin, the teammate has its own identity and operates autonomously within set boundaries. The firm's announcement comes amid broader interest in and awe of AI agents — since Meta released Muse, its new personal AI agent, earlier this month, it's climbed to the top of Apple's US App Store.

Hedge funds have long pursued machine learning and have more recently poured huge sums into AI systems to help their investment teams. Millennium has also invested heavily in AI talent, launching an AI track for its internship this year and building an AI laboratory to develop cutting-edge technology and lure top technical minds, per Bloomberg. The firm also announced a partnership with Anthropic last month to co-build an "AI-powered digital risk analyst," which will support its risk management teams and is part of an effort to "deploy Anthropic's most advanced models across teams."

As expensive as the AI efforts can be, Millennium is in a good spot to finance them. On October 1, it will close on $22 billion in new capital commitments.

• Governments are seeking to set restrictions on popular tech products, from social media to AI.
• However well-intentioned, the restrictions likely won't stop people from using the tech.
• Just look at the lessons the US learned after outlawing alcohol for stretch more than a century ago.

When the US outlawed alcohol in 1920, people found ways to keep drinking — whether by sneaking into secret speakeasies in dingy basements or by obtaining dubious prescriptions for medicinal booze.

By 1933, when lawmakers finally legalized hooch, drinking levels were comparable (about 70%) to what they were before.

The failure of prohibition to make a meaningful dent in the country's imbibing habits is instructive as tech companies and governments increasingly seek to impose limits on the use of ubiquitous software, such as social media and AI for kids and teens.

In recent months, Meta announced a slew of restrictions for teen users of Instagram and Facebook, New York City effectively banned AI in its public school classrooms, and frontier AI companies have called on the federal government to set limits on their development.

The measures, however well-intentioned, could have the unintended consequence of pushing people toward unregulated platforms without proper safeguards.

"My concern about the bans is that people will go underground," said Uttara Ananthakrishnan, a professor of technology at the University of Washington. "When they go underground online, bad things happen — really terrible things happen."

The data, Ananthakrishnan told Business Insider, shows that flat-out bans on tech people already use rarely work. Now, with VPNs — software allowing an online user to spoof their true location — they've become pretty fruitless.

The limits of social media bans

Ananthakrishnan said her research has broad applications, particularly in the context of social media bans.

"People are just going to find other ways to get around it," she said of the bans. "Enforcing it is incredibly difficult."

In a study published in April, Ananthakrishnan and her colleagues found that when adult websites implemented age-verification checks, people rapidly migrated to websites without them.

"This pattern suggests that age verification laws did not trigger systematic diversion of users from adopting or exiting platforms toward non-adopting ones," the study says.

Last year, Australia barred kids under 16 from using any social media apps. As Business Insider's Katie Notopoulos pointed out, the measure is hardly having any effect.

The parental-control tool company Qustodio said in a report that within a few months, teens and preteens got right back online at levels that matched pre-ban numbers, largely by entering incorrect ages or using VPNs.

And last month, Meta settled a landmark social media addiction lawsuit by agreeing to implement a number of restrictions on teen users of Instagram and Facebook.

Among the restrictions are a two-hour daily daytime limit for under-18s, a default nighttime block for young users, and a notification block during school hours.

Though Meta tried to entice other short-form video providers, such as TikTok and YouTube, to enact similar restrictions, that hasn't happened so far.

Ananthakrishnan said that without standardized, industry-wide regulation, teen users will quickly migrate to other platforms that offer unrestricted access.

"The network is very easy to shift," she said.

AI restrictions

New York City and Los Angeles, which have the first- and second-largest school districts in the country, both announced restrictions on the use of generative AI in the classroom last month.

In New York's case, students in the city's elementary and middle schools will have no access to generative AI programs on school-owned hardware. High school students will take two "AI literacy courses" each year and will have access to AI programs in limited educational contexts.

Some have said that the moratorium will have New York's students using AI in an unmonitored, unregulated fashion.

"They're using AI outside of the education space," Keri Rodrigues, the founding president of the National Parents Union, told Business Insider.

"They are going to use all kinds of means to go around the system, but also overconsume. But also overconsume in secret," she said. "Every kid I know — and I have five sons — has ChatGPT on their phone."

Rodrigues said she'd like to see AI policies that teach students how to use the tools responsibly and safely.

The lesson of prohibition

There are some limits to the analogy here. For one, it wasn't exactly possible, a century ago, to use a VPN to purchase alcohol from, say, an online distributor. The workarounds were often explicitly criminal and required daring and complex operations to smuggle booze from Canada or the Caribbean while avoiding undercover federal agents.

Also, amending the constitution to outlaw alcohol amounts to a far more thorough and drastic step than if, for example, individual manufacturers imposed limits to the alcoholic content of certain beverages.

That being said, there are some very poignant parallels. Alcohol, like Instagram, was and is a ubiquitous part of American life — though people are drinking less alcohol now.

Social media levels, however, remain persistently high. A 2023 Gallup study found that 51% of US teens spend more than 4 hours per day on social media.

And now, with easier methods of circumvention, tech companies and governments might find that restrictions are easier said than done.

"Once you get a taste of whatever your convenience is, and if the convenience is taken away from you, people will find ways to get back to that convenience," Ananthakrishnan said.

• Meta poached MongoDB's CEO and made clear it wants a much bigger piece of the AI pie.
• Software's AI fears are back, but investors are being more selective than during the SaaSpocalypse.
• Sign up for First Trade, Business Insider's daily markets newsletter.

Another week. Another move from Meta that flips the AI trade upside down.

Last week, Meta's agentic-AI push minted new winners in the chipmaking space and sent its own shares up sharply, cementing the company as a frontrunner in consumer AI.

But CEO Mark Zuckerberg doesn't seem content to stop there — he also wants to disrupt enterprise software. Meta made the bold move of poaching MongoDB president and CEO CJ Desai right before the company's investor day, appointing him to run the new enterprise platform.

MongoDB got crushed, falling 18% on the day. Perhaps more surprising was that Meta fell 5%. Investors may be weighing the risk of pursuing consumer and enterprise AI at once, although profit-taking after a huge post-Muse rally was likely also a factor.

The more consequential moves came in software. Investors quickly began repricing a familiar fear: Meta's arrival could further threaten incumbents already under pressure, although losses did ease as the day progressed.

It turns out Zuck had another ace up his sleeve all along, and Muse was just one part of the master plan. The question now is how it will balance both initiatives — and convince investors that each is on track.

Ambitions beyond consumer AI

The Muse push made a ton of sense to investors because Meta already has a massive built-in audience across its multiple apps. If they can convert even a sliver of those people into Muse users, the upside is immense.

Breaking into enterprise software and convincing clients to use Meta AI for their businesses is another matter entirely. The company needs credibility if people are going to use its tech stack throughout their operations.

That's where Desai comes in. He's someone with existing relationships that can get people to start using Meta software and, most importantly, stick with it long term.

Software isn't one trade anymore

So what does this mean for the market? Monday's stock recovery in the space shows it's not as simple as "software go down." This isn't SaaSpocalypse 2.0. There's some nuance to the winners and losers.

The software companies most at risk are those whose products handle repetitive tasks: workflow tools, customer-service platforms, and developer software. If Meta's agents can do more of that work directly, those companies could lose pricing power.

The most defensible area may be the firms that make agents usable: data-management providers like Snowflake, plus cybersecurity companies such as CrowdStrike and Palo Alto Networks. Meta can build the interface, but it still needs help connecting it safely to a business's systems.

Ultimately, Meta is trying to own both ends of the AI boom. And, as usual, investors are pricing outcomes on the fly. Expect more volatility across the space as other gigantic companies unveil their own transformational ambitions.

"If you can't raise a dog, you cannot run an office."

That was the message Nico Laqua, cofounder and CEO of buzzy insurance startup Corgi, sent to staff in the New York office about one of the company's short-legged mascots. A vet had told the employees that the puppy, Hamilton, should not be outside because he wasn't fully vaccinated, so they were using pee pads in the office.

After Laqua, based in San Francisco, heard that workers were not walking Hamilton, he laid down the law in a July 19 missive.

"Buying pee pads of any sort for Hamilton is a fireable office [sic], on the spot," he wrote in the message. Those who refused to take the corgi to the street should book a one-on-one. "The best case scenario for this meeting will be a PIP," he said, using shorthand for the disciplinary process known as a performance improvement plan.

It's hard to imagine this kind of C-suite drama playing out at Liberty Mutual, Berkshire Hathaway, or other giants of the buttoned-up commercial underwriting industry. But Corgi is decidedly not your granddaddy's insurance company.

The CEO is loudly online, bragging his way across X while communicating with his staff on the messaging app Telegram. A squad of young women dubbed "Corgi girls" relentlessly promotes the company and their own hot takes on social media. It recently embarked on a massive expansion of a curious side hustle: a chain of 24-hour Corgi cafes.

The startup is constantly announcing new products and ventures; it has an exchange-traded funds arm, a plaintiff's legal firm, and file-sharing software, but the core product is insurance. Corgi markets its startup business insurance in the forefront, and offers the usual options, too: insurance for your apartment, your truck, your bar.

"We want to be a one-stop shop," Jeremy Eisemann, senior vice president, told Business Insider.

The frenetic pace carries over to its financials. It's raised over $431 million; in May, it announced a $106 million round three weeks after a $160 million unicorn round. The company is now valued at $5 billion, according to a person familiar with the matter.

Haden Kirkpatrick, a longtime insurance tech advisor and investor who is not involved in Corgi's business, said he'd never seen an insurance company raise money like it. "The companies that hold risk in this space — that actually have a premium on paper that they have to reserve against — don't come anywhere near that," he said.

The raises impressed Kirkpatrick, who said he understands why investors were excited. Corgi had that Silicon Valley ethos. "If you're a technology company, you're at warp speed," he said. "If you're an insurer, you're crawling."

Business Insider spoke with current and former employees, analysts, investors, and prospective customers about Corgi's trajectory, and reviewed dozens of documents. One thing was clear: Corgi is speedrunning in a sleepy sector. Some say it's disrupting a business that needs a shake-up. Others caution it's been tried before: Lemonade, Hippo, and Root vowed in the 2010s to upend the status quo, then fizzled.

"It's not so much that they failed and went away," said Nathan Golia, a senior analyst at Celent. "It's that the idea that insurance was ripe to be picked off by a tech company turned out to not really be correct."

Neither of Corgi's cofounders had insurance experience. They knew how to make Roblox games.

Laqua, now 26, met Emily Yuan during her sophomore year at Stanford, before she dropped out. They created the social app Picnic and scaled it into the gaming studio Basket Entertainment, earning glossy magazine stories and a spot on the Forbes "30 Under 30" list. After they got into the startup accelerator Y Combinator, Laqua and Yuan obtained insurance licenses and launched Corgi in 2024.

The new company quickly spent $30 million on YRIG, an Alabama-based tenant liability insurance company, according to people familiar with the matter. "How do you buy a $30M insurance carrier with just $500K from YC?" Jared Friedman, a YC partner, said in an investor's blog post. "It seems impossible. And yet, Nico and Emily pulled it off."

YRIG isn't a traditional insurance agency but a risk retention group. Corgi also operates through the Arizona-based risk retention group TRRG. That means if Corgi were to fail, policyholders under YRIG and TRRG wouldn't be guaranteed payouts by the state. Corgi said that the company operates through many insurers, including admitted or fully licensed carriers, and that RRGs are "industry standard for certain types of risks."

YRIG is also unrated, and Corgi said it does not plan to seek a rating for it. Corgi's Eisemann said the unrated status didn't diminish customer trust and that Corgi has more than enough funds to meet operational requirements. "We will be there when a covered claim is filed," he said. "The money is on the balance sheet. We have it sitting in the bank."

Corgi grew its ranks quickly, opening eight offices, including San Francisco, Atlanta, Salt Lake City, Chicago, Dallas, New York, and London. It now has about 400 employees.

Former employees of Corgi, which has a seven-day workweek, describe a hyper-competitive culture. Throughout the day, Laqua uses Telegram to rank workers and offices against each other on who has performed the most insurance demos, messages seen by Business Insider show. Laqua told Business Insider that it has a "bot that tracks revenue activity, like most companies do."

The sales team's approach isn't always appreciated, based on social media posts by prospective customers complaining about an onslaught of calls, texts, and even FaceTime outreach.

Ari Ramdial, the San Francisco-based founder of the stablecoin wallet Guap, thought Corgi's coverage might be a fit for his startup. Seconds after entering his information on the company's site in May, his phone began blowing up with calls and texts from employees — a startling deluge even for a sales-based industry.

"At one point, I was getting two calls from two different Corgi numbers simultaneously," he told Business Insider.

He responded to one email, saying it was "incredibly annoying," and requested that they "please stop calling me repeatedly." A Corgi rep responded by apologizing and offering an "apology gift," which he did not accept. Asked about Ramdial's complaint, Corgi told Business Insider that "customers appreciate responsiveness," and that those who ask not to be contacted are put on a no-call list.

Corgi's sales pitches to potential clients have also been the subject of litigation. AppFolio, a leading property management system, sued Corgi's property insurance subsidiary, Beagle, in December, alleging that it deceptively told customers it was associated with AppFolio, then took over customers' accounts and switched their insurance. Beagle is "named for a pet, but its business runs like a parasite," AppFolio wrote in the suit.

Beagle denied many of AppFolio's claims and countersued, alleging that AppFolio engaged in a "systematic campaign to destroy Beagle's business," which AppFolio denied. The lawsuit is ongoing.

In February, a Beagle account executive posted in the company's Telegram chat that another employee told a prospective customer that they "directly integrate" with AppFolio. "We do not," the account executive wrote in a message viewed by Business Insider. "We are literally getting sued bc of this."

A business development representative chimed in to ask whether they could say they "work with" AppFolio. "That's what I say," he said. Beagle CEO Anthony D'Angelo responded that it's a "Soft integration."

Beagle said in a statement to Business Insider that it is not an AppFolio partner and that "soft integration" referred to scheduled reports that Beagle clients set up in their AppFolio accounts.

Since its founding, Corgi has filed at least five other lawsuits, including one involving a former executive who accused the company of age discrimination.

The legal dispute began when the company fired Grant Bigler, a vice president, last October. He responded with a demand letter to the company that called it "youth obsessed," according to court documents. Bigler didn't sue, but in a preemptive strike, Beagle went to court and asked for a declaratory judgment that his dismissal was not related to age. Beagle amended the suit, which is pending, to include an allegation of theft of trade secrets, which Bigler denies.

Current and former employees told Business Insider that the company, like many startups, runs young. Dominique Sutton, 30, said she felt like a "grandma" during her eight months working in operations at Corgi. While she loved her time there — she called it exhilarating — she said it was a culture shock adjusting to a company where "everyone there is between 18 and 21." Corgi said it did not have that data on the age of its workforce.

The employees were smart, Sutton said, but they acted like kids. She recounted them saying, "Let's get a dart board. Let's get a gong. Let's get the most asinine things a young teenage boy would need."

As Business Insider previously reported, a "little red book" purporting to contain Corgi's philosophy included the declaration that "boomers are slow and bad" and that "the best athletes, warriors, and chess grandmasters are 19-30 years old." Corgi said the book was created and given to them by an outside investor and that it does not endorse its full contents.

One upside of a younger employee base is social media savvy, and Corgi is highly online. The company posts leadership lessons on LinkedIn and newsletters on Substack. Internally, staffers are ranked on a leaderboard for social media engagement, with prizes promised for top performers, Telegram messages show. Corgi told Business Insider it didn't create the leaderboard and that it has no incentives for social media posting beyond one competition, where the winner got to "help up a new office."

The most eye-catching of the company's social media presence: the Corgi girls, as the twitterati call them. These employees post less about the nitty-gritty of insurance and more about their lives — everything from travel to matcha. There are cleavage-baring selfies and the occasional double entendre. "1 girl / 14 guys / 1 founder house," one posted with a revealing photo. "wonder where I've seen that film before."

The San Francisco slang dictionary, SF-isms, defines Corgi girls as "the attractive, extremely-online women Corgi (the YC S24 AI insurance unicorn) hires for marketing, growth, and events. Their hiring announcements flooded tech X and started a debate about hot people as B2B marketing strategy." SF-isms author Mehul Agarwal told Business Insider the insurance firm's quest for virality seems like a smart business move. When he moved offices, the real estate agent told him, "Get Corgi."

"He used Corgi instead of saying, 'Get insurance,'" Agarwal said. "That is a clear indication that something here is working."

As the company's head of community, Brooke LeBlanc leads the march of the Corgi girls. She set off the most recent company kerfuffle by posting to X her list of non-negotiables in a man, which included "provider/protector mentality" and being "post-economic."

The September 17 post went viral, and the New York Post called it "cringeworthy." Corgi staffers rushed to LeBlanc's defense, and Laqua rang the "cancel culture" alarm. When one X wag opined that Corgi girls had "Hooterized themselves," an employee called the commenter a "fake looking bitch" who "has her tits out" and later, in a defense of free speech, wrote that she would not "attempt to police" anyone who used the N-word. (Corgi said it does not "endorse or support any use of profanity, slurs, or unprofessional conduct.") In the midst of the furor, LeBlanc did the unthinkable for a Corgi girl: She deactivated her X account.

Sutton doesn't get all the fuss about Corgi girls. "I love how it ruffles everyone's feathers who are not used to being comfortable around women," she said. Eisemann said he was too busy to follow the drama but believes in "freedom of speech." Others are turned off — Guap's Ramdial said that he wants an insurance provider to be "good and boring."

Boring is not in the CEO's lexicon. In one tweet, Laqua is telling convicted fraudster Martin Shkreli that he and his friends wrote him letters in prison when they were teens; Laqua later said he did not write any letters and deleted the post. In another post, Laqua is trashing well-liked investor Sheel Mohnot over a mild observation on X. In one May post, he touts Corgi's then-$1.3 billion valuation. The next day, he's alerting his followers that there's a 50% off sale at the Corgi Cafe.

The cafes started with a single location in San Francisco, then spread throughout the city. This summer brought the announcement that it planned to open 100 around the world within a year. The all-night coffee shops have sparked memes and drawn one small protest. One recently made headlines after being shuttered for operating without a permit, and the executive behind the expansion has reportedly left the company.

Caribou Honig, a longtime insurance tech investor who has no money in Corgi, said he was reminded of the Capital One Cafes. Corgi is trying to turn insurance into a "high-engagement experience," he said — but wondered if that's what its core business customers want.

I visited New York's Corgi Cafe five days after it opened in August. Barista Mirta Atemis, who said she was making a healthy $23 an hour, told me customers were mostly Corgi employees who work upstairs and the occasional curious local. "I think people don't know whether we're open or not," she said.

The back of the shop was a coding zone. There are phone booths for calls, and small circular tables for locked-in work sessions. Anish Basu, the 43-year-old founder of Worthy, said he heard about the cafe on X and came by, excited by the idea of meeting other founders while working.

There was only one other customer that day — a grad student, not a founder. "Give it a couple months," Basu said. When I returned last week, there were at least a dozen people working in the back room on a weekday afternoon.

One thing not seen on either visit: Corgi's famous corgis. The company is named after Laqua and Yuan's first office pet, Trudy, and it now has other corgis in various offices.

There's a long history of friendly insurance animals: the Geico gecko, the LiMu Emu. Tom Amico, who created the Aflac duck with Eric David, because the company name sounded like "quack," is puzzled by Corgi's choice of mascot.

"Don't ask me what corgis, or Queen Elizabeth's dogs, have to do with startup insurance," he said. "I guess it's cute."

Well, not always. Having a company pet is all fun and games — until it poops on the floor, as one former employee said Trudy often did. The company later set up a Telegram bot to ask employees to walk her.

Laqua may run a billion-dollar business with hundreds of employees, but the corgis can be top of mind. In early July — while Corgi revved up for Y Combinator events to recruit interns, according to one insider — Laqua told the New York office he wanted puppy Hamilton flown to San Francisco, a Telegram message shows. A few days later, Laqua had another demand about the 6-week-old: "No pee pads or he will be trained to pee inside on every carpet forever."

Staffers were worried about the pup going outside or on a plane. "Can we take hammy to the vet and maybe get his shots early or just check up if hes okay to fly?" one asked in the chat on July 10. Another staffer asked them to delete the message: "Nico gona tweak if he hearrs pee pad."

A week later, the tension was higher. "I just heard Nico say that there will be no NY office if Hamilton isn't out here by Wednesday morning," one employee wrote on Telegram on July 19. Another reported Laqua said that he's "shutting down nyc" if the dog didn't make the trip.

After Laqua insisted on outdoor potty breaks, a worker wrote that "the vet specifically told me he shouldn't be going outside until his last round of shots." Another staffer suggested: "Make sure to tell the airport people that hes 4 months and has his shots" — although X posts and messages indicated he was about 8 weeks and not fully vaccinated.

Corgi told Business Insider that it "takes the care and wellbeing of its office dogs very seriously" and expects the employees around them to care for them. Separately, Laqua said that Hamilton was 8 to 9 weeks old at the time of the trip and that vaccines are not required on domestic flights, which aligns with major carriers' written policies.

Hamilton did make the trip, snoozing in his carrier on the flight. He survived San Francisco, though the visit wasn't without a hiccup.

"Found Hamilton chewing on a zyn during last call," one staffer wrote on Telegram. "Pls make sure floors are clear."

Additional reporting by Jack Newsham

Do you have a tip about Corgi? Contact reporter Henry Chandonnet via email at henrychandonnet@proton.me or on Signal at henrychand.30. Use a non-work email address and device; here's our guide to sharing information securely.

• Today's tip to get better at work: Become a more confident leader by showing humility.
• Jamie Dimon advises leaders to encourage open communication and be open to criticism.
• For decades, he's left every JPMorgan board meeting so directors can speak freely without him present.

Get better at: Leading

Today's move

Jamie Dimon, the CEO of JPMorgan Chase, said insecurity ruins CEOs and other leaders.

So instead of acting like you have all the answers, actively aim to create a space where others feel comfortable weighing in and disagreeing.

Why it works

Dimon said that insecure leaders make their reports uncomfortable about bringing them bad news or challenging their ideas. Eventually, this causes leaders to create a bubble around themselves and they become disconnected from what's going on inside their companies.

Encouraging others to share their perspective, on the other hand, signals humility and confidence in a leader.

Try it today

The next time you meet with your team, end the conversation by asking your peers, "What am I missing?"

Then just listen to what they have to say.

Dimon said that he isn't offended by complaints and that a secure leader values critical feedback.

The bottom line

Dimon actively encourages his employees to speak up. He said that for decades, he has left every JPMorgan board meeting so directors can talk without him. When executives disagree with him, he has them make their argument directly to the board.

Source

This insight is drawn from Business Insider's reporting of Dimon's take on insecure CEOs from his commentary on "The Master Investor Podcast with Wilfred Frost." Read the full story about Dimon's comments from Business Insider's Georgia Hennessy here

It also includes Business Insider's reporting on how to spot a fragile ego, based on Dimon's comments and advice from leadership experts. Read the full story in Business Insider's Sarah Needleman here.

What do you consider to be the signs of a secure leader? Tell us what you think of Jamie's leadership advice in the comments below.

• MKBHD didn't think Tesla would release a $30,000 Cybercab to Americans by the end of 2026.
• The YouTuber made a bet: If he's wrong, he's shaving his head on camera.
• Over the weekend, Elon Musk talked about a new timeline for the Cybercab.

Tesla CEO Elon Musk offered a fresh Cybercab launch timeline — and MKBHD's barber can breathe a little easier.

In an interview with the state-run China Global Television Network, released on Saturday, Musk outlined when he expects the company's two-seat, steering-wheel-free vehicle to launch more commercial rides.

The Cybercab "will soon be operating commercially in Florida, in Nevada, and a couple of other states," he said. "And, probably, by the middle of next year, it will be operating commercially in California."

That does not technically settle Marques Brownlee's famous Cybercab bet. The techie YouTuber, known as MKBHD, said in 2024 that he would shave his head on camera if Tesla matched its promise to sell the steering-wheel-and-side-mirror-less vehicle to regular buyers for $30,000 before the end of 2026.

For years, the internet has run wild with that boisterous wager, especially on X, where users have posted AI-generated videos of Musk shaving the YouTuber's head for the better part of two years.

Even Musk has chimed in, writing on X after the Cybercab's commercial launch in Texas that MKBHD "really wants to shave his head."

Musk's new timeline — especially his admission that California won't see a commercial robotaxi release until next year — adds a new wrinkle for those hoping to see a bare-headed MKBHD. The company also has to resolve an NHTSA audit query into its self-certification that the vehicle meets federal safety standards.

In an early September video, MKBHD reiterated that he was pretty confident that he would win the bet, saying that Tesla was still several steps away from putting the aggressively priced vehicle in Americans' driveways.

"My hair is safe, for now," he said in the video. "And honestly, I feel pretty good about it staying good for the rest of the year."

Representatives for MKBHD and Tesla didn't respond to requests for comment from Business Insider.

• Instinct creator Noah Shinn said that trust is key to keeping users engaged with the viral AI assistant.
• Instinct has data showing that users who share a credit card number are much more likely to stay.
• Have you used Instinct or a similar AI assistant?

Instinct's creator, Noah Shinn, says he cares deeply about how his users feel.

Shinn said that Instinct, the viral AI assistant startup, already has data showing that if a user connects one piece of sensitive information (like a credit card number or account password), then they're more likely to stick around.

"There's an interesting data flywheel or chicken and egg problem here, which is the more data that you give to it, the more proactive it can be, the more sympathetic to your situation it can be, the more that it has to just act and be more useful to you, right?" Shinn told "The Invest Like The Best" podcast in an episode posted on Monday. "What we find in the data is actually that it takes several weeks to build trust."

Instinct found that "there's a 40% chance that the user has shared a personal credit card" within their first three weeks. In that case, or if a user shares similar sensitive data, Shinn said there's an 80% retention rate.

"Through that process over the first, I guess, three weeks, the user learns to build trust," he said. "And then from there, it's just a snowballing effect."

Part of building user trust, Shinn said, also requires drawing a clear red line on how Instinct will try to monetize itself as it expands.

"We don't want Instinct to influence the user's behavior in a way that is not aligned with what the user wants," he said.

Shinn said Instinct doesn't want its users to be the product, calling out established tech platforms that have done the opposite.

"I'm talking the major platforms, whether that's Google or TikTok or Instagram or Snapchat or so many others, where you have this platform, and it's free for the users, but then there's so many instances in which paid ads are pushed to the user, and they try to convince the user to then purchase," he said. "And by the very nature that they do purchase, it now becomes this game of these brands paying to convince users to purchase items that they may or may not want."

One of the reasons Shinn is against such ads is that he said it's not a stretch to expect a day when AI assistants like Instinct are smarter than their users. Exploiting that power and losing trust in the process would be a bad move.

"I think it would be a very dangerous world if Instinct were influencing the user's behavior to purchase something that they don't want to purchase," he said, "or to subscribe to something that they don't want, and using its intelligence to be able to convince them."

• Many people are making two big mistakes in their retirement accounts, Empower's Marta Norton says.
• She pointed to how many workers need to increase contributions to get their employers' full 401(k) match.
• Some Americans are also too exposed to stocks as opposed to stable-value investments, she added.

Many Americans have no idea they're making two big mistakes with their retirement savings, one investment strategist says.

Marta Norton, the chief investment strategist at Empower, one of the nation's largest 401(k) providers, opined on the biggest mistakes she saw when it came to Americans investing for retirement. On the client side, there are two errors that consistently stick out: not investing enough in your 401(k), and investing too much into stocks, she said, speaking recently to Bloomberg Radio.

First, many Americans could increase 401(k) contributions, Norton said, noting that some workers weren't contributing enough to get their company's full matching contribution.

A 2015 study conducted by the investment advisory Financial Engines estimated that Americans leave a collective $24 billion in unclaimed 401(k) company matches every year.

"Everyone points out that's free money," Norton said of the employer match. "I think that's probably the most obvious mistake."

Second, many Americans are taking a do-it-yourself approach to retirement investing — and are getting too much exposure to stocks, rather than to assets like cash and stable-value investments, Norton said.

Most retirement accounts are heavily exposed to stocks, a strategy to build long-term wealth. But investing too heavily in equities could work against older Americans, who are closer to exiting the workforce and are thought to be more vulnerable if stocks were to see a significant correction.

A 2023 Fidelity report found that more than a quarter of baby boomers had an "aggressive" allocation to stocks in their retirement accounts, meaning their portfolios were weighted more heavily toward stocks than the target allocation Fidelity recommended for their age.

"If you're thinking about what is the hidden risk for some people who are kind of investing on their own in retirement, it might be that they're taking on a bit too much equity than is appropriate for their particular situation," she added.

More attention has been shed recently on the dangers of overinvesting in stocks, particularly as movements like Financial Independence, Retire Early garner more online pushback. In the FIRE community, the risk of having to withdraw retirement savings during a market downturn — locking in the losses on investment — is known as the sequence of returns risk.

• Data centers aren't just powering your new favorite agent; they're also a key construction driver.
• The AI buildout is set to become the largest capital expenditure project in US history, dwarfing railroads and highways.
• Overall construction spending has fallen year-over-year, while data center spending has skyrocketed.

Data centers are powering your chat ramblings — and the growth in the American construction industry.

AI buildouts are set to become the largest capital expenditure project in US history, per a Brookings Institution analysis, outpacing spending on highways and railroads as a share of GDP.

That massive spending and construction boom needs workers to — literally — prop it up. And blue-collar employment for data centers is booming, with the number of nonresidential specialty trade contractors — workers like electricians in non-residential buildings, which includes data centers— growing by 86,000 in the past year.

"I think it's fair to say that a very large portion of construction labor demand right now is being driven by increasing data center construction activity," said Zach Fritz, an economist at Associated Builders and Contractors.

The American Institute of Architects found, based on multiple forecasts of nonresidential construction, that data center construction spending is set to climb more than other kinds and account for about 8% of nonresidential building spending by 2027.

But the sustainability of that jobs boom could be precarious. Construction spending on the whole has fallen 3.8% year-over-year as of July, according to the Census Bureau, and KPMG finds that data centers are a "structural growth driver" for spending. Year-over-year spending on data centers has grown by just over 57%, a sharp contrast to both residential and non-residential private spending growth since early 2025.

For now, the data center boom is good news for construction workers and larger industry spending.

Guy Berger, a senior fellow at the research organization Burning Glass Institute, said one concern is if the data center boom cycle ends up being shorter than anticipated: "And all of a sudden, we trained a lot of electricians, people like that, to do stuff that is no longer needed," Berger said.

Data center workers are in demand

Job postings on Indeed for data center-related roles have skyrocketed, up almost 130% from about two years ago, while overall postings are down compared with the same period. About half of the data-center demand this year came from the IT infrastructure, operations, and support occupational group and the installation and maintenance occupational group.

A job-posting analysis from Lightcast, a job market intelligence company, showed a rise in demand for some data-center-related jobs. Based on growth in average monthly job postings from January to August 2022 and the same period this year, postings for electrical engineers within the utilities sector have increased 34%, and 22% for project management specialists. Within commercial and institutional building construction, there's been a rise in demand for construction managers, project management specialists, cost estimators, and civil engineers.

And so, in the short term, the data center boom will likely remain a workhouse for the construction industry. What's more unclear is the longer-term prognosis, especially since AI infrastructure is still reliant on a slice of hardcore spenders to keep it afloat.

"The next few years, it's stable. It's going to keep growing. We're going to keep building data centers," Fritz said. But rising community opposition could eventually limit the rate of growth, he said, or technological advancements might require smaller footprints.

"There's all kinds of things that could happen here," he said. "They're not going to happen anytime soon."

It's unclear how permanent data center jobs will ultimately be, especially after data center construction, but listings on Indeed at least suggest firms are advertising them as long-term. Aubrey Woessner, an economist at Indeed Hiring Lab, said only about two in every 1,000 data center postings are temporary.

"'Permanent' is a label the employer chooses, and there's reason to think it may overstate how durable these roles really are: Local incentive agreements often require a project to create a set number of permanent positions, which gives employers a reason to post that way regardless of how long a given build-out job actually lasts," Woessner said.

Are you in a data center-related job? Reach out to these reporters to share at mhoff@businessinsider.com and jkaplan@businessinsider.com.

• Krystal Pulaski switched her intended career path after soccer injuries led her to physical therapy.
• Pulaski's honesty about her knee surgery during job interviews helped her secure multiple offers.
• She chose a physical therapy assistant job in San Marcos, Texas, prioritizing work-life balance.

This as-told-to essay is based on a conversation with Krystal Pulaski, a 24-year-old physical therapy assistant who lives in Greensburg, Pennsylvania. It's been edited for length and clarity.

I earned a BS in kinesiology in May 2023, and then I graduated with an Associate of Applied Science degree in January 2025.

After passing the National Physical Therapy Examination (NPTE) in April 2025, I began applying for physical therapy assistant positions in Texas. I was applying to any job I felt like I'd be a good fit for. The offers started rolling in about three weeks later.

I received multiple job offers to work as a physical therapy assistant. Among those included was one from a skilled nursing facility that paid around $30 an hour, and one from an outpatient facility that paid around $26 an hour.

I accepted the job paying $30 an hour in San Marcos, Texas, and started my new job in May 2025.

As I was figuring out my career path, healthcare wasn't always part of my plan

I was a soccer player, and I planned to join the military right after college. That changed after I went through physical therapy myself following a series of injuries. The initial injury that led me to work with a physical therapist happened while I was at a semi-pro soccer training in the summer of 2020. I was doing a drill that tore my meniscus in my knee.

Being on the patient side showed me how much impact a physical therapist can have. Physical therapy helped me regain strength, function, confidence, and independence. I realized I wanted a career where I could give others that same hope and support. That experience ultimately inspired me to become a PT assistant.

I applied for physical therapy jobs while preparing for my fourth major knee surgery

My doctors had told me that after the surgery, I wouldn't be able to walk for six weeks. The timing made my job search feel uncertain because I saw all my classmates getting jobs immediately and felt behind. I asked myself, "What if all these jobs are gone by the time my knee is done healing?"

I tried to decide whether to wait and apply after I was done recovering and able to walk again, or take the risk and apply to see if a job was out there that would work with me during my recovery.

I decided I'd apply and see if the job would be willing to wait for me to walk again, or find a job that would work with me while I was recovering.

I chose to be open about my situation during interviews

Somewhere in the middle of my interviews, I brought up, "I'm going to be having major knee surgery this month." My interview for one of the jobs was a week before the surgery was scheduled, my interview for the second job was the day before my surgery, and my interview for the third job was the Monday after my surgery.

I told the interviewers that I was willing to do what I could and would work with them while I wasn't ambulatory, but that I'd need to be non-weight bearing for six weeks. There were a couple of hiring managers who said they needed someone who could start immediately, which I understood, but there were others who said, "If you want to start working a couple of weeks after your surgery, that's fine — or we can wait until you're able to walk again."

I feel like being candid about my situation helped me get multiple offers because if you're open and honest, people are more willing to work with you. That turned out to be true, since after sharing my personal situation, I received multiple offers from employers willing to accommodate my recovery or wait until I was fully healed.

The job I chose offered the best work-life balance

The hours are very flexible. I come in at 6:30 a.m., and once I'm finished with my patient caseload, I can leave. At the latest, sometimes I'm there until 4 p.m.

I also chose this job because the environment is vibrant and outgoing, and the complexity of the patients keeps me learning, as I adapt exercises and treatment techniques within each patient's plan of care.

When I was considering other job offers, one of the biggest factors for me was the commute. Some of the positions would've required a much longer drive into areas with extremely heavy traffic. I knew that spending a lot of time commuting each day would've hurt my work-life balance and overall quality of life.

The other offers were still appealing for other reasons, but the position I accepted stood out because it offered more flexibility in addition to higher pay and a manageable commute.

Starting my career with a personal situation that was challenging to share with employers was a little nerve-racking

It also confirmed that I chose the right path. Physical therapy is a world where you never stop learning. Each patient brings their own challenges, and that's what keeps the job interesting and pushes me to grow.

About a year into my first role, I started looking for a new job to be closer to my boyfriend in Pennsylvania. I started as a physical therapy assistant at an outpatient clinic in Greensburg, Pennsylvania, in August.

The transition has been busy, but I'm excited to keep building my skills, forming connections with my patients, and starting this next chapter of building my career.

• Georgia Tech robotics professor Animesh Garg is joining Microagi as chief research officer.
• More robotics academics are launching and joining startups amid the physical AI boom.
• Microagi has raised $55 million and aims to deploy 10 million robots to customers in five years.

For decades, the biggest advances in robotics came out of university labs. Now, as billions pour into physical AI, the researchers behind that work are launching and joining startups.

The latest to recruit from academia is the German company Microagi. The startup has hired Animesh Garg, a Georgia Tech professor and former Nvidia researcher, as chief research officer, the company exclusively told Business Insider.

Garg will lead its work on post-training, adapting general-purpose robotics models into systems capable of performing specific tasks. He is bringing seven members of his research group with him through Microagi's acquisition of a startup the group had been building in stealth. Garg and Microagi declined to disclose the terms of the deal.

Microagi raised $55 million this summer in the largest seed round raised by a German startup, Business Insider previously reported. The company works with manufacturers in the automotive, logistics, and food industries to train robots for factory jobs. It uses data collected from workers to adapt general-purpose robotics models to each customer’s operations.

Garg's move comes as money pours into robotics and physical AI, fueled by advances in AI, falling hardware costs, labor shortages, and pressure to automate and reshore manufacturing.

The boom is also drawing more robotics researchers to startups. Carnegie Mellon professors Deepak Pathak and Abhinav Gupta cofounded Skild AI, which has raised more than $2 billion to build a general-purpose AI "brain" for robots. Stanford professor Chelsea Finn and UC Berkeley professor Sergey Levine cofounded Physical Intelligence, another startup developing general-purpose robot models that has raised more than $1 billion. This summer, MIT professor Russ Tedrake brought Walden Robotics out of stealth.

Garg said he decided to join Microagi because robotics is rapidly moving from impressive lab demonstrations toward large-scale commercial deployment.

"The technology has crossed a threshold," Garg told Business Insider. "We have gotten to a point where we can create sustainable businesses."

Garg has spent more than 15 years researching how robots learn. He received his Ph.D. from UC Berkeley, completed a postdoc at Stanford, and held faculty positions at the University of Toronto and Georgia Tech. He also served as chief scientific officer at humanoid robot maker Apptronik from 2024 to 2025.

At Nvidia, Garg worked on simulation technology that was eventually incorporated into Isaac Lab, Nvidia's open-source platform for training robots virtually. His more recent research has focused on teaching robots from human demonstrations and helping them improve through experience.

That work matches Microagi's approach. The startup collects data from people performing physical tasks, including workers wearing cameras and sensor-equipped gloves, and uses it to train robots.

CEO Bercan Kilic told Business Insider that Microagi wants to help deploy at least 10 million robots for customers over the next five years.

Have a tip? Contact Rya Jetha via email at rjetha@businessinsider.com or Signal at rjetha.07. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

• An ex-OpenAI researcher built Jev, a popular new AI model that takes a different approach than LLMs.
• The model is made for code, makes decisions in milliseconds, and includes probabilities in answers.
• Developers at a Saturday hackathon told Business Insider what they're building with the new tool.

When a new twist on AI called Jev arrived this month, San Francisco's tech scene responded as it does: lots of hype, and then a hackathon.

Jev is a new and viral AI model built by the startup TypeSafe AI, and developers are flocking to try it out. Jev's launch video from mid-September has racked up almost 40 million views on X, throwing founder Diogo Almeida into the spotlight.

Notably, Jev marks a break from large language models (LLMs) at OpenAI and Anthropic. Rather than spitting out text, Jev makes decisions. It's also dirt cheap and incredibly fast. An ex-OpenAI researcher, Almeida has said that the focus on chatbots doesn't do enough to automate rote work.

I attended a Jev hackathon on Saturday at the startup CodeRabbit's office to find out what the hype is all about. The event became something of a coming-out party for Jev. Some developers say they have already woven Jev into their products and feel that TypeSafe has delivered a true breakthrough. Sourabh Mane, a design engineer at CodeRabbit, called it a "breath of fresh air."

Venture capitalists are buzzing around Jev as well. The Information reports that some VCs have offered to invest in TypeSafe at a $10 billion valuation.

How Jev differs from chatbots like ChatGPT

Jev takes in questions and rapidly responds with structured answers that include probability estimates of accuracy. It works best when engineers break down their questions into the most specific possible pieces, Allie Laabs, TypeSafe's head of developer relations, told the crowd at the start of the hackathon.

Laabs told me that, with Jev, her startup hopes to advance a new school of thought in an industry dominated by LLMs.

"What if this path that we've been so horse-blinders on, the LLM path, what if that's not the only path?" she said.

Laabs gave the developers some other pointers, too. Jev works well for classifying — it could mow through customer support queries and slot them into buckets, or moderate a fast-moving online chat feed. She hopes that Jev's speed and low cost will help some developers build tools they'd previously thought were too ambitious.

Developers sprint to inject Jev into their code

Fueled on croissants and coffee, the developers were thrilled to get going.

Cofounder Jalaja Kurubarahalli said that if Jev can sort questions more quickly, it could reduce the lag in the voice tools that her startup, Quintess AI, builds for maintenance workers. Hitesh Viswasam, a robotics engineer, suggested that a model like Jev could let robots respond to heat or danger at human-like speeds.

Some attendees had already improved their code with Jev. Developer Lawrence Liang showed me the personal AI agent he'd built for himself. He said this was his second Jev hackathon of the week and that he'd built the model into 80% of his agent's code infrastructure. Jev figures out much more quickly than an LLM which tools are needed for any given query.

"It almost blows your mind," Liang said.

Engineer Avram Cheaney said that adding Jev to his AI search product for eBay, called JunkDrawer.ai, had made it "drastically cheaper and drastically faster" than using just LLMs. Those few seconds of speedup are particularly important for consumer products, where people have become used to instantaneous loading.

"This is what's been missing," Cheaney said.

TypeSafe peddles a new way to think about AI

It's been a hectic week at TypeSafe. Jev's launch catapulted the company from anonymity to a closely watched AI player. Its Discord server now has more than 100,000 members.

"It is definitely completely nuts," Laabs said, adding that the early success is something "very few people ever get to experience in their lives."

Laabs said she's been delighted to see that people have quickly begun to understand what Jev can do.

Before, she was worried because Jev doesn't align precisely with how AI is widely understood, where a text box spits out code or detailed answers to questions.

"A software engineer has to put it in something; it isn't anything until that," Laabs said. She's found that the explanation helps: "Then people go, 'Ohhh!'"

Have a tip? Contact this reporter via email at scouncil@businessinsider.com, or over text, Signal, Telegram, or WhatsApp at 415-757-8198. Use a personal email address, a nonwork WiFi network, and a nonwork device; here's our guide to sharing information securely.

• Nvidia robotics alumni are increasingly striking out as founders.
• AI advancements and a flood of venture capital are fueling the wave.
• Nvidia could still win as its alumni build on the ecosystem it created.

Nvidia helped build the foundation for physical AI. Now, many of its veterans are launching the next wave of robotics startups.

The chip giant has poured billions into robotics, and Nvidia CEO Jensen Huang helped popularize the term "physical AI, " which enables machines to interact with the real world.

As with AI agents and other applications, Nvidia's strategy is to build underlying technology — the chips that power robots and the software used to train them. That ecosystem has become fertile ground for founders.

In recent years, some who helped build the foundations of robotics inside Nvidia "are spinning out to build companies of their own," said Amulya Vishwanath, who left the company last year to start Techable Ventures, a venture firm focused on physical AI.

Nvidia's former vice president of AI research, Sanja Fidler, left in July to found Veeda AI alongside other ex-Nvidians. The startup builds simulated worlds to train robots — an extension of the research Fidler worked on at Nvidia. Former Nvidia researcher Joel Jang founded Dream Labs, another startup working in that industry, the Information reported.

Nvidia alums are attacking different areas in robotics, from software for deploying robots in warehouses to those that perform physical tasks for businesses.

Nvidia has invested in some, including Flexion Robotics and Dyna Robotics.

Flexion CEO and cofounder Nikita Rudin said that Nvidia maintains close ties with the startup, holding biweekly meetings and access to early tools for testing.

"We're still doing the same thing we were doing before," Rudin said. "We were the beta testers — and now again, just from the outside."

Why Nvidia robotics talent is striking out

Rudin and his cofounder, David Hoeller, left Nvidia after becoming convinced robotics technology had reached an inflection point and was ready for deployment.

"We were not necessarily convinced that staying at Nvidia was the fastest way to actually go and deliver that," Hoeller said.

Former Nvidia robotics researcher Arsalan Mousavian cited a similar motivation. He described Nvidia as a "hardware-first" company centered on AI chips, and he wanted to be "laser focused" on building useful robots. Mousavian's startup, founded last year, remains in stealth.

Siddhant Haldar, a former Nvidia robotics intern who considered joining the company before founding Index Robotics, said the field has long been held back by a lack of real-world training data. Now that powerful AI models can pick up some of the slack and robotics venture funding is booming, he said it was the right time to strike out.

Physical AI companies raised $33.4 billion globally in the first half of 2026, already surpassing the $28.7 billion raised in all of last year, according to PitchBook.

Startups can also take risks and deploy imperfect robots because large companies have "more to lose," Haldar said.

Nvidia isn't alone in producing a new crop of robotics founders. Former Google DeepMind engineers launched Reimagine Robotics and Generalist, while Waymo veterans have gone on to build Bedrock Robotics, as researchers from fields like autonomous driving are pouring into the space.

Nvidia's growing startup diaspora creates a cycle: The more startups that emerge, the more Nvidia stands to benefit by selling them the infrastructure needed to build.

Have a tip? Contact this reporter via email at gweiss@businessinsider.com or Signal at @geoffweiss.25. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

When I was a young lad spending too much time on the internet, I was obsessed with fonts.

I read blog posts on sites like Typographica about the finer points of kerning. I read white papers from Microsoft about the readability of Calibri compared to Arial. I copied and pasted hundreds of pages' worth of books and essays from Project Gutenberg into Microsoft Word and formatted them to my liking. Among them was a theory of great fiction and taste by Henry James, published in 1884, which I later recast in Cambria and printed out so I could read it on the bus.

James believed that greatness could be defined by its elements, but that taste comes from experience. "Some people, for excellent reasons, don't like to read about carpenters; others, for reasons even better, don't like to read about courtesans." It was futile, he said, to impose what "people ought to like or dislike."

Back then, I was trying to discern what I liked and why I thought those things were good. Fifteen years later, I'm still writing (nothing as good as "Portrait as a Lady"), and don't care much about fonts anymore (I wrote this story in Google Docs using Arial). But I realized that the process of figuring out the music and clothes you like — developing taste — is not so different from figuring out which words to use and how to arrange them on a page. It takes time, detours, and mistakes. Taste takes life to acquire.

Now comes a company with $18.5 million in venture capital funding that supposes: What if AI could do all that for you?

The thesis of Taste Labs is that taste is a form of pattern recognition and, therefore, ripe for artificial intelligence to master.

The problem with AI today, says Taste Labs CEO Thais Castello Branco, is that it produces things that might be technically excellent, but it lacks the guidance to understand which parts of its training data are good and which parts are junk. AI-generated outputs are, as a result, statistically average. At worst, you get images of terrifying-looking food. At best, you get a bland LinkedIn post or a soulless actress. In other words, you get slop.

AI researchers have long recognized this problem. The technology excels in "objective" domains like coding. It struggles in subjective areas like design, creative writing, and fashion, where a right answer could be dependent on context, culture, and differences of opinion. To solve that, Taste Labs is creating taxonomies of quality. Castello Branco tells me that the company is breaking down designs — starting with product websites, slide decks, and brochures, before eventually branching out to higher-order creations — into measurable elements, and then creating the right data to train frontier models to become better at judging and measuring quality.

The central question Taste Labs aims to answer, says Castello Branco, is: What are the criteria that go into defining what is great? The central question Taste Labs raises is: Do we want AI to define what is great?

The idea that an AI algorithm can determine good taste might seem hubristic for a company based in Silicon Valley, the community that brought us Allbirds, digital tokens of unshaven apes smoking cigars, and the Metaverse. Who the hell are these people to think that taste is a problem to be solved by computers? And how dare they think computers can solve it?

Castello Branco, 29, is quick to admit her own taste can be "basic." She says her favorite movies are Christopher Nolan films and the Brazilian crime drama "City of God" (she grew up in São Paulo) — a staple of the IMDb Top 250 list. After graduating from the University of Chicago, she developed products for Heinz Kraft, including pickles, desserts, plant-based food, and a "burger sauce." Later, she cofounded Exa, a company that operates a search engine for AI agents, before launching Taste Labs last October.

Castello Branco's solution is to outsource taste to a network of about 1,000 "Tastemakers" — contractors in creative fields who train artificial intelligence models on what is good, and what is not. As the logic goes, their aggregate preferences improve upon the entire-internet slurry of data that's been used to train AI models so far. "Very high quality curated preference data," she says, helps to improve "subjective domains" like creative writing and architecture.

Indexing and systematizing all forms of human creativity is a tall order. Taste Labs is starting with visual design, a field already in flux with Figma and the rapid rise of Claude Code and Cursor. The task involves breaking down great design into its subcomponents and making sure that AI models can execute each of them well. The models have to visualize the code, understanding how people would interact with it, select appropriate typefaces and colors, and figure out the right spacing and alignment. Castello Branco's panel of Tastemaker experts, she says, could largely agree on the merits, or lack thereof, for each element. Claude Code might think that the recipe website you asked it to design should have a lot of introductory text; Taste Labs knows the recipe should be near the top. Eventually, the company plans to branch out into other fields, like copywriting, literary fiction, music, and cinematography.

Business is booming, Castello Branco and investors say. The company has 30 full-time employees. Veronica Orellana, a partner at CRV, one of Taste Labs' lead investors, tells me that the startup already has "tens of millions in revenue." The steady financial success and clear vision mean Taste Labs doesn't need to raise more money anytime soon.

Taste Labs is working with "most of the major frontier labs" on various projects, Castello Branco says. She declines to say which.

"Sometimes they come to us with a very broad problem," Castello Branco says. "Something like, 'Hey, evaluate our model for web games capabilities, figure out what's broken about it, and then suggest data that could fix those problems.'"

The Tastemakers do this in a variety of ways. For some projects, they create their own designs and then write explanations of their particular creative choices. In others, they write evaluations of different AI outputs. Some Tastemakers evaluate the work of other Tastemakers, further refining the data.

Whitney Bolin, an art director who's worked for companies like Adobe and Adidas, tells me she's worked on tasks that involve "breaking down design and style into things that are kind of measurable and observable, like typography, motion, color, textures — things that could be subjective and giving them a grade or evaluating them as someone with taste."

"There's a lot of thought and consideration put into each evaluation," she said. "It's not as simple as just rating out of 10."

Castello Branco was inspired to start Taste Labs after fiddling with widely available AI models and finding that the results for creative work — websites, graphics, that sort of thing — sucked. She looked around and saw piles and piles of slop.

"I really didn't want to live in that world," she continues. "I was like, I don't want this. I want to live in a world of creativity and craft and quality. And I feel like that's really missing."

I know what she means. Between interviews with Castello Branco, I was using AI to vibe code a new personal website to showcase my writing. Google's Stitch web design tool had the bright idea of including a snippet of one of my stories on a yellow legal notepad, a nod to my days covering courtroom trials. But the execution seemed wrong, even when I used Business Insider's Codex account, an expense I can now justify by including this sentence in this article I'm writing for work. The yellow that Codex selected was garish instead of mellow, and the font was bizarrely florid. I tried to prompt it so that the edges of the paper would curl when I hovered over it, giving the website a tactile visual susurrus of rustling through the notepad's thin pages. But I just couldn't get it to look right. These were more issues of baseline quality control than personal taste.

The first product Taste Labs launched is a Brand API. You can plug in a brand's website, and it can extract and systematize the visual components into a set of rules. That rulebook, in turn, can be plugged into graphic design tools to help companies create new graphics while remaining within the brand's guidelines. A company like Coca-Cola could hire an ad agency and give it access to an API that'll ensure any artificially generated ads use the correct shades of red and their new font without constant prompt reminders. Another tool allows users to search for cohesive brand systems. If a pillowcase design entrepreneur told the API, "I want a dreamy brand," it would produce a color scheme that evokes clouds and sunset colors.

For the past few years, AI companies have been proclaiming that their products will sweep away the drudgery of work and give us the freedom to focus on human creativity. Taste Labs' grandest ambitions, to automate key components of creativity, go beyond that. The logical endpoint for Taste Labs seemed like the end of culture itself: to get AI to do all the things humans are supposed to do.

Castello Branco reassured me that Taste Labs isn't looking to replace humans. With AI, you no longer need to learn how to code to make a website, or play an instrument, or use specialized software to make a song. Taste Labs could supercharge all that.

"I don't think it's productive to assume that every person in the world that is not an expert in this domain will take the time to become an expert," Castello Branco says. "They shouldn't."

Bolin, the art director, wrestled with the question when she answered an online ad to join the Tastemakers network and learned she would be working for an AI company. She ultimately believes that AI can never replace "the emotional intention and value that a human can add."

Like me with my digital garish yellow notepad, Bolin found herself frustrated when she sometimes lacked the technical abilities to express her creative vision. Taste Labs can create tools that can "allow people to get past that technical hump," she says.

"There's always going to be an evolution," she says. "In the '90s, Photoshop and Adobe Illustrator came out, and a lot of people were saying, 'Oh no, designers are going to be replaced.' Everything was done by hand."

Most people won't take the time to master website design or architecture, but Taste Labs could help them execute their visions at a higher level, rather than stumble around in the dark with Claude prompts. Maybe I could get back into typography again. Maybe I could design my own fonts with AI, and have Taste Labs figure out the kerning instead of wondering what my life would be like had I gotten an internship with Tobias Frere-Jones instead of Interview magazine.

Castello Branco shares her own hesitations about traversing into more creative domains. She recognized that developing AI that could "guarantee a minimum bar of quality" was related to, but different than, catering to an individual's taste. For the creative people who push culture's vanguard, Taste Labs would help them bring their ideas to life. "If you want to create a very specific aesthetic, my dream is that you can prompt that and get exactly that very well executed exactly to your taste," Castello Branco says.

I pushed her. Let's say we're decades in the future. AI — and Taste Labs — reaches a quality level beyond anyone's wildest dreams. Christopher Nolan is dead. You've seen every one of his movies, except for "Following," available only on MGM+, which you're never going to sign up for. Would you ask Taste Labs to make a new Nolan movie? What would that look like?

Mimicking a particular artist's work is the wrong mission, she said. It's more about learning from Nolan's greatness to produce something that would be great in its own right.

"It's not imitating Christopher Nolan, but can they create a movie that is as good as Christopher Nolan or directionally closer to Christopher Nolan?" she said. "That is much more the goal."

Castello Branco's perspective was not so different from what Henry James believed. Taste is subjective, and it could not — and should not — be imposed from above. And besides, taste is an ever-evolving issue. It's not something that has a solution. One purpose of having a large Tastemaker network, Castello Branco said, is to try to create something that represents "the actual diversity of taste that exists in the world."

Something kept nagging at me. As James wrote, taste comes from experience. Experience, he said, was "an immense sensibility, a kind of huge spider-web, of the finest silken threads, suspended in the chamber of consciousness and catching every air-borne particle in its tissue" that could be harnessed for creative work.

That experience — the living, the tinkering around, the trial, the error — is exactly what Castello Branco was trying to cut through. How can you figure out what you like if AI is doing all the hard work for you?

I was inspired by my conversations with Castello Branco. I wanted to make my own font. I started chatting with ChatGPT. I wanted a take on Minion Pro, an Adobe font I admired for its clarity and elegance. Some of the characters, particularly the lowercase "y," always seemed oddly stubby to me, so I decided to make a version of the font that had a bit more panache when it came to the descenders.

ChatGPT's first attempt was terrible. If it had Taste Labs integration, I'm sure it would have been better. I kept suggesting refinements, trying to get the font to a place that would satisfy me. The "y" ended up different than what I originally intended, but something that I liked well enough. Was it any good? I wasn't sure. Maybe I could ask ChatGPT.

• AI agents can make purchases for you, but not everyone is ready for it.
• A survey found 53% of respondents were uncomfortable with an AI agent making a purchase for them.
• Trusting agents enough to take action could take time and require changes to check out systems.

The moment many have waited for — where AI can actually do your bidding, not just tell you how to do it — may finally be here, but that doesn't mean consumers are ready for it.

Makers of personal AI agents like Meta's Muse are touting their ability to handle shopping tasks from beginning to end: researching items, comparing prices, and checking out on your behalf. While many consumers are open to having the AI agent handle the first two steps in that process, trusting it with the final step could prove more difficult.

A recent survey conducted by YouGov for ACI Worldwide, a global software and payments company, asked over 3,000 adults from the US and UK how they felt about using AI agents for shopping related to fashion. The survey found 53% of respondents said they would not allow an AI assistant to make purchases on their behalf under any circumstances. Only 7% said they would allow agents to make purchases without their approval.

Other data suggests that skepticism isn't limited to shopping for clothes and shoes. Accenture's Consumer Pulse Research report, which surveyed over 25,000 consumers worldwide, found that while 74% of consumers would trust a personal AI agent more than their best friend to make a purchase on their behalf, only 9% were open to fully autonomous shopping. About one in three said they'd let AI decide what to buy, but that they would personally make the payment.

As the personal AI agent moment appears to have arrived, the data suggests there's still a fundamental disconnect between wanting an AI agent to help with shopping and trusting it enough to fully hand over the reins.

"Right now, it's a trust issue," Dan Coates, director of merchant product management at ACI Worldwide, said. "People have gotten comfortable letting AI help them with discovery, comparison, and decision-making, but they still want to be the ones to pull the trigger."

Coates said it's still the wild west when it comes to agents making purchases on people's behalf, but that a purchasing framework for agents could help consumers feel more comfortable with the idea, adding, "think of it as an agentic bill of rights or rules of the road."

Some payment companies are already working on such programs, like Mastercard Agent Pay and Visa Intelligent Commerce. These agent frameworks, he said, could lay out rules for agents, consumers, merchants, and banks, adding transparency and security to the process.

Phil Bruno, chief strategy and growth officer at ACI Worldwide, said that while the capabilities of AI agents have improved, it only takes a few negative stories to impact user trust.

Early users of personal AI agents have already reported some issues. A tech reviews YouTuber said Muse gave his address to a potential buyer on Facebook Marketplace and arranged a pickup without his knowledge. A Meta Superintelligence Labs staffer said on X that in similar cases Muse was "following direct instructions and correctly asked for permission." A representative for the company previously told Business Insider it attempted to contact the YouTuber but did not get a response.

While Muse can make purchases on users' behalf, Meta says the agent will always ask for approval first.

Bruno said trust would likely build up gradually as users experiment with agents, starting with the "tinkerers" who are trying it out now and eventually with a broader set of early adopters. He said that testing will likely start with merchants that users already trust and are already comfortable with one-click ordering, like Amazon.

He also said that asking whether AI agents are ready to take over purchasing decisions isn't the right question.

"It's not that it's got to be ready, but it's got to be better. And how much better?" he said. "Is it appreciably better in a way that now I actually want this to occur?"

We're interested in how people across different generations are using — or not using — AI agents. Tell us about your experience in the survey below.

• Jalen Brunson said his parents taught him to be realistic about what he could take on during high school.
• The New York Knicks player ultimately had to decide what he was willing to give up.
• He said his friends played an unexpected role in helping him stay on track with his goals.

New York Knicks' Jalen Brunson says the right environment made it easy for him to put basketball first in high school.

In a joint appearance with his mother on the "Things No One Tells You" podcast, Brunson said the support he received from those around him helped lay the foundation for his success.

Brunson said his parents taught him that he only had the time and energy to excel at two of three things in high school: grades, basketball, or a social life.

"School was clearly wasn't an option to say no to, so it was really between basketball and my social life — and I chose basketball," Brunson told podcast host Lindsay Czarniak.

He said his days revolved around keeping up with his homework and putting in the time the sport required.

"And then after that, I could be a kid," Brunson said. "I could do what I want."

Having friends who understood his priorities helped him stay on track, he said.

"I had a great group of friends who knew my goals, who knew what I had to do before I could do anything," Brunson said. "They never pressured me into skipping a day to go hang out, whatever. They made sure I did what I had to do."

Those friends are still in his life today, he said.

"It was really important for me to understand that my surroundings were just as important as my decisions," Brunson added.

In June, Brunson led the Knicks to their first NBA championship in 53 years. He was also named Finals MVP after scoring 45 points in the team's 94-90 Game 5 win over the San Antonio Spurs.

The NBA player has previously spoken about how the people close to him helped him stay disciplined.

Brunson said his parents taught him from a young age that there are no shortcuts in life. In a 2019 interview with ESPN, he recalled how, during his freshman year, they began leaving notes around the house reminding him that "the magic's in the work."

"They'd put it on the bathroom mirror, on the garage door before I left for school, on the refrigerator. They put it everywhere they knew I'd look. They'd send me a picture of it. They wanted me to understand as long as you work hard, eventually things are going to go well," he said.

In an appearance on the "Come and Talk 2 Me" podcast in 2024, Brunson said his team at Villanova helped him develop the discipline that college basketball demanded.

"I had good habits, but I needed better ones because college is different. So being able to be held accountable by the older guys on the team and by Coach Wright and stuff, it definitely helped me," Brunson said.

• AI companies are getting tired of employees passing off chatbot responses as their own.
• Leaders at tech startups such as Clay and Figma began sharing their companies' AI writing policies.
• One company said that it's asking employees to use a steak emoji to flag non-human work.

For the past three years, the em dash has been a popular tell for identifying AI writing. Now, AI startups want their employees to self-disclose.

On Monday, Varun Anand, the cofounder of AI sales and marketing startup Clay, shared his company's recently rolled-out AI writing policy in an X post.

The guidelines, which were shared on LinkedIn last month, do not ban employees from using chatbots to brainstorm and draft their writing, but ask them to take responsibility for everything they share with colleagues.

"It is your responsibility to make sure that the entire document is representative of your own thoughts before you share it," the document read. "If a reviewer asks, 'What did you mean by this line?', it's not acceptable to reply with 'Oh sorry, AI wrote that, just ignore it.'"

Anand wrote that the policy began with Clay’s engineering team before spreading companywide. It argues that writing is itself a form of thinking and that sending a long AI-generated document after typing a short prompt offloads the work of making sense of it onto those who have to read it.

Following Anand's post, other tech leaders responded with their own versions of how employees are being asked to flag AI writing.

Figma CEO Dylan Field wrote in an X post that his company has a straightforward approach. The design software company’s policy requires workers who use AI in their writing process to add “written with AI” at the top.

Bobby Ghoshal, the founder of AI shopping startup Dupe.com, wrote that the company’s policy is to add a steak emoji when workers paste AI-written copy. The symbol is meant to show that the employee is acting as a “meat proxy” for machine-generated words, he wrote in his post on X on Monday.

“Varun is touching on a live wire inside every company right now,” Ghoshal said, referring to Anand's post.

In a reply, an employee of AI research lab Exa said that AI writing at the company is banned.

"No external comms (tweets/linkedIn post/and more) can be AI written, and everything needs to be hand written," Ishan Goswami wrote. "If you don't take the time to write it, why should anyone take the time to read it."

The companies are responding to a complaint that is becoming increasingly common in workplaces: AI is making it effortless to produce text, but employees are spending more time deciphering it.

While the Clay AI policy prompted a conversation about how AI writing is being flagged within companies, broader efforts to disclose AI-generated content began earlier this summer.

In August, Anthropic began embedding an imperceptible watermark in text generated by its new Claude models, after signing on to the EU AI Act’s transparency code. The watermark doesn't change the text's meaning or readability, but it travels with the text when copied and pasted and "may persist through some editing," the AI lab said.

AI companies had already established AI disclosures for more visibly risky media. Google’s SynthID can embed invisible watermarks in AI-generated images, video, audio, and text. OpenAI has added Content Credentials to images from DALL-E 3, ImageGen, and Sora, and this year began adding SynthID watermarks to supported images and audio generated through its tools.

• A new Ukrainian report says an interceptor drone took down a target at 264 mph.
• That's fast enough to cross the length of a football field in less than a second.
• The target was a Geran-5, said to fly at speeds of up to 373 mph.

The drone battle to take down Russia's latest Gerans looks to be moving at speeds faster than Formula 1 cars or bullet trains.

The latest Ukrainian report to highlight such extreme speeds came from the Serhiy Prytula Charity Foundation, a volunteer organization that supplies Ukraine's military, on Monday.

The foundation posted thermal footage shot from a drone approaching another uncrewed system above the clouds. The shape of the target closely resembles the Geran-5, a Russian one-way attack drone that resembles and shares many characteristics with cruise missiles.

The camera closes the distance on the Geran-5, almost reaching it before the footage cuts off — a typical indicator of a successful or near-successful interception.

"The drone was destroyed while flying at 425 kmph, at an altitude of more than 3 km," the foundation wrote. That's about 264 mph, fast enough to cross the length of a football field in less than a second.

The foundation said the footage was filmed by a crew from Ukraine's 1020th Anti-Aircraft Missile and Artillery Regiment, which is based in the central Ukrainian region of Poltava.

The unit had used a new interceptor drone that was "designed specifically" for jet-powered drones, it added.

Ukrainian intelligence has said that the Geran-5, described as a Russian version of the Iranian Karrar drone, can reach a top speed of 373 mph, making it the fastest among its one-way attack drones.

Though the reported interception was much slower at 264 mph, the speed is a new combat achievement for Ukrainian interceptor drones, which are designed to be cheaper and mass-producible.

The fastest speed ever clocked by a Formula 1 car, for example, was 246.9 mph in a land-speed record attempt by a modified Honda F1 car. The reported interception was also faster than Japan's world-famous bullet train, the Shinkansen, which reaches speeds of 200 mph.

Ukrainian officials reported similar speed achievements late last year, though these weren't recorded in combat.

In December, Mykhailo Fedorov, who was Ukraine's digital transformation minister at the time, said an interceptor drone had reached a speed of 400 kmph, or about 248 mph, with a special motor.

Ukrainian interceptor drones, which rose to prominence as a counter to the Russian Shahed in late 2024, have typically been propeller-driven quadcopters shaped like bullets.

But recent designs emerging to battle turbojet Shaheds and attack drones have begun to feature jet engines, too. It's unclear which type was used in the interception reported on Monday.

Ukraine's defense ministry said that it has been field-testing at least four types of interceptor drones, including a winged drone that's launched via catapult.

The Wild Hornets, a major interceptor manufacturer, has also been releasing footage of its newest interceptor, the Sting-S, chasing down jet-powered drones. Last week, the firm told Business Insider that the Sting-S is a modified quadcopter-type interceptor.

• PwC surveyed 50,000 workers globally about their professional "hopes and fears."
• Its findings split workers into four distinct groups based on how and how much they're using AI.
• This divergence is creating a "two-speed" workforce, said PwC's global workforce leader.

Is AI turning you into a "frontrunner" or are you working in the "engine room?"

A huge survey carried out by consulting giant PwC found that AI is allowing a small group of workers to pull away from the rest of the workforce and dividing employees into four distinct categories.

The Big Four firm published its "2026 Global Workforce Hopes and Fears Survey" on Tuesday. It surveyed almost 50,000 workers across 48 countries and 29 sectors between May and June 2026.

AI use is rising, the survey found. 64% of workers said they had used AI at work in the past 12 months, up 10 percentage points from a year earlier, and the share using generative AI every day rose from 14% to 22%.

But the survey revealed a widening divide in how workers are getting the tools and opportunities to benefit from AI.

Based on the responses, PwC identified a group it called "front-runners," which make up 14% of those surveyed.

The front-runners have in-demand skills and reported getting strong benefits from AI in their work. Just over half the group use generative AI daily, while nearly 80% said they have access to learning and development resources.

Then there's what PwC calls the "engine room" workers — 56% of workers who PwC characterizes as the core of organizations’ day-to-day delivery.

Just a small number of these workers say they use generative AI daily, and fewer than 40% say they have access to learning and development resources.

"They're not getting the same access to learning. They're not getting the opportunity to innovate. They are not getting to use AI in a meaningful way," said Peter Brown, PwC's global workforce leader.

PwC's third category is "AI insurgents," who make up around a fifth of staff. Their skills are less in demand than those of front-runners, but they're ambitious in how they use AI.

Finally, there are the "indispensables," who have scarce skills that are highly valued by employers.

The outcome of this divergence showed up in findings on job security, workers' confidence in asking for a promotion, trust in managers, and skills development — all of which were lower for the engine room cohort.

The divergence between the groups is creating a "two-speed" workforce, said Brown.

The survey does not establish the reasons behind workers' experiences, and PwC’s worker categories are partly based on respondents’ reported experiences with AI and on how in demand they believe their skills are.

Still, the findings point to a practical challenge for employers spending heavily on the technology and seeking ROI: buying tools is not the same as changing work across an organization.

Ignoring the divide between employees risks missing out on productivity, revenue per employee, job security, and satisfaction, Brown said: "The danger is that you actually could render a big chunk of your workforce largely irrelevant in the world of work."

PwC, one of the Big Four accounting and consulting firms, itself employs over 360,000 people globally and has raced to build internal chatbots, upskill staff, and test AI in client work.

The firm pitches itself as "client zero" for AI transformation while advising companies on how to make the same shift. While the firm has made some mistakes along the way, Brown said PwC has embraced AI, which has brought positive changes to the work and the tools employees have access to.

The changes have included rewriting PwC's training agenda around 30 core skills: 15 AI-centric and 15 human-centric, a change introduced in February 2026. The firm has also reduced the number of offices its entry-level US consultants can join, in a bid to improve community and learning opportunities.

Brown said companies should be transparent about why they are using AI and the outcomes they hope to achieve. “Workers aren’t expecting leaders to, I think, sugarcoat everything,” he said. “They just want to understand what’s going on."

"Leaders who display those kind of characteristics and others tend to be the ones that seem to be bringing the workforce along with them," he added.

For employers, Brown said, widening access to AI is not an either-or choice between backing the front-runners and investing in everyone else.

The results show that "there's an enormous amount of value there that can be tapped" if leaders can get more of the workforce to lean in, he said.

Whether you're a social media doomscroller or a frequent video caller, unlimited data is definitely an enticing perk for most smartphone users. Forget jumping on shoddy airport WiFi or rationing monthly allowances — the right carrier can grant you limitless data, eliminating the need to worry about any of that. With that in mind, we've gathered the best unlimited data plans for your cell phone below, based on our extensive testing and research.

As the particular coverage area of a carrier should be your primary consideration, we've selected the best unlimited data plans from each major carrier in the US: Verizon's Unlimited Plus plan, T-Mobile's Experience More 2.0 plan, and AT&T's Extra 2.0 plan. These tiers are not the carriers' cheapest or most expensive options; they're happy mediums offering the best value for most people with full network access and still liberal mobile hot spot data.

We've also listed Mint Mobile's Unlimited 12-month plan and Visible's Plus Pro annual plan as budget options. While both offer unrivaled long-term value at especially low cost, especially for fewer lines, note that premium unlimited data plans from major carriers can still be surprisingly affordable, and perhaps easier to manage, with multiple lines.

Our top picks for the best unlimited data plans

Best from Verizon: Unlimited Plus Plan - See at Verizon

Best from T-Mobile: Experience More 2.0 Plan - See at T-Mobile

Best from AT&T: Extra 2.0 Plan - See at AT&T

Best budget: Mint Mobile Unlimited 12-Month Plan - See at Mint Mobile

Best budget alternative: Visible Plus Pro 12-Month plan - See at Visible

Best Verizon unlimited data plan

Verizon's Unlimited Plus plan offers the best balance of features, data connectivity, and value. We like the Unlimited Plus plan's truly unlimited premium data, with no data caps and speeds not artificially reduced by network traffic. It also offers access to Verizon's fastest 5G network.

The Unlimited Plus plan is pricey, starting at $80/month for a single line with the auto-pay discount. However, the monthly cost per line decreases as you add more lines, making it more affordable to add friends and family to the plan. For four lines, it can go as low as $42.50/month, and Verizon is, accordingly, our top choice among the best family cell phone plans.

For the high-end price, however, the Unlimited Plus plan gives you the peace of mind that you'll always get the fastest speeds available in your coverage area.

To be sure, Verizon's cheaper Unlimited Welcome plan, which starts at $65/month for a single line (with an auto-pay discount), also includes truly unlimited data. With four lines, the monthly cost per line lowers to $27.50. However, you don't have access to Verizon's fastest 5G network, and your line is subject to deprioritization, meaning data speeds can be artificially throttled when the network is congested. It also doesn't include the features below.

The Unlimited Plus plan includes 30GB of mobile hot spot data monthly. That's plenty of data to connect a laptop or tablet to browse the web, send emails, watch videos, or scroll through social media.

Assess Verizon's coverage map and check with local friends, neighbors, or social media like Nextdoor to ensure adequate coverage.

Best T-Mobile unlimited data plan

T-Mobile's Experience More 2.0 offers an excellent balance of features, data connectivity, and value. We like the plan's truly unlimited premium data without hidden data caps or deprioritization. However, speeds can still be naturally affected if the network is bandwidth-constrained. The plan also offers access to T-Mobile's fastest 5G data speeds.

Starting at $85/month for a single line (with an auto-pay discount), the Experience More plan is very pricey for one or two lines. Yet, it's well-suited for those who can add friends and family to their plan. The total monthly cost for three lines is $140, which translates to about $47 per line. Add a fourth line, and the monthly cost reduces to $42.50 a month per line, which is an excellent value.

However, thanks to T-Mobile's premium data plan, single-line users can rest assured they'll get unlimited data at the fastest speeds available in their coverage area. There's no data deprioritization here.

T-Mobile's cheaper Essentials Saver 2.0 plan costs $50/month for a single line. It includes a generous 50GB of premium data with access to T-Mobile's fastest 5G network. Lines that use over 50GB of data are then deprioritized. Essentials customers may also notice slower speeds than other customers during periods of congestion, per T-Mobile, and the plan doesn't include the features listed below.

The plan includes 50GB of mobile hot spot data, which easily suffices for web browsing and sending emails on laptops and tablets, and even allows plenty of headroom for video streaming.

Texting and WiFi connectivity on flights are also included, and video streaming is possible when it's available, depending on the airline and flight. T-Mobile also lets you upgrade phones every two years, compared to the typical three-year upgrade window.

T-Mobile also offers a partial discount on Apple TV Plus and a free ad-supported Netflix Standard plan with the Experience More 2.0 plan.

To ensure coverage in your area, check T-Mobile's coverage map and survey local sources like friends, neighbors, and online forums.

Best AT&T unlimited data plan

AT&T's Extra 2.0 Plan offers a great balance of features, data connectivity, and value. The plan includes 100GB of premium data before deprioritization, which is less generous than the truly unlimited premium data you get with Verizon and T-Mobile's comparable plans. However, 100GB should easily suffice for most people, and you still get unlimited data after 100GB, though it's subject to speed throttling when the network is congested.

Starting at $70/month for a single line with the auto-pay discount, the Extra 2.0 plan is slightly more affordable than Verizon and T-Mobile's equivalent offerings. The plan becomes more affordable with each additional line, with the monthly price per line dropping to $40 at four lines.

AT&T's Extra 2.0 plan comes with 50GB of mobile hot spot data, which is more than enough data to browse the web and send emails from a laptop, with even some headroom for video streaming.

Check AT&T's coverage map and local sources to ensure coverage.

Best budget unlimited data plan

Mint Mobile's Unlimited 12-month plan for $30/month (or $15/month with a limited-time discount) offers incredible value for truly unlimited data. As a budget-friendly mobile virtual network operator (MVNO), Mint Mobile is understandably lighter on extra features than the premium plans from major carriers. Still, it may offer all that most people need in an unlimited plan.

Mint Mobile runs on the network of its parent company, T-Mobile, and its data is subject to deprioritization behind T-Mobile's higher-paying customers. While Mint Mobile's unlimited plan is truly unlimited, going beyond 35GB of monthly data can deprioritize speeds for users on the carrier's lower data tiers, but only during periods of congestion.

The Unlimited plan from Mint Mobile includes 10GB of mobile hot spot data, which should be plenty for browsing the web and sending emails from a laptop, and even the occasional video stream, with an emphasis on occasional.

Mint Mobile doesn't include access to high-band 5G networks, also known as mmWave 5G networks, but most people may not notice. The carrier advertises typical speeds between 79 and 357 Mbps, which is more than fast enough for most common phone activities.

Mint Mobile's plans are prepaid in three-, six-, or 12-month increments, with lower monthly prices for longer intervals. Thanks to its contract-free, subscription-style format, it's easy to avoid unnecessary fees and add-ons. It's our top pick among the best cheap cell phone plans for eligible users.

Mint Mobile runs on T-Mobile's network, so consider its coverage map and local T-Mobile sources to determine whether the network works in your area. For further assessment, see our Mint Mobile review and Mint Mobile vs. T-Mobile guide. If you're already looking to join, check out our guide to the best Mint Mobile deals to get the best discount on your membership.

Best budget alternative

If T-Mobile's service doesn't perform well in your area, then our other favorite budget MVNO option is Visible, which uses Verizon towers instead. It offers truly unlimited, prioritized data on all of its plans, the top tier being Visible Plus Pro, which costs $45 a month (or $39 a month with the promo code SAVE6). It's on sale sporadically throughout the year, making deals like the current 50% off annual plan promotion a good time to sign up.

Owned by our top pick, Verizon, it uses the provider's towers to provide similarly reliable service at a considerably lower price, in exchange for fewer bells and whistles.

That doesn't mean that Visible Plus Pro doesn't come with sweet perks, though. For that $45 monthly starting price, it comes with truly unlimited premium data, automatic high-risk spam call blocking, included smartwatch service, limitless hotspot data, and more.

One of the best features of Visible Plus Pro is its built-in travel accommodations. The provider's Global Pass is the equivalent of a travel eSIM, allowing users to use data, make calls, and send texts while abroad without racking up a hefty bill. Plus Pro automatically grants customers 24 Global Pass days every year, free to use immediately. It's a simple in-app setting to toggle it on and off.

If you don't travel as much, or don't care too much about "premium" data, the MVNO's two cheaper plans, Visible and Visible Plus, start at $25 and $25 a month, respectively.

The main limitation of Visible as an overall provider is its lack of family or group plans. Each user needs their own account and to sign in; there are no multi-user plans. This may not be a major problem for couples or friend groups, but for families with kids, it makes account management and payment unnecessarily difficult.

If that's not a problem for you, check out Visible's coverage map to ensure your region has reliable service available. You can also learn more about our experiences testing it in our Visible review and Mint Mobile vs. Visible guide. Also, keep tabs on our guide to the best Visible deals to find the right sign-up discount.

How we test unlimited data plans

We test unlimited data plans at scale as part of our broader testing for the best cell phone plans, using standard criteria and special considerations.

With preset data limits for nominal unlimited plans, or the lack thereof on truly unlimited plans, our primary focus is testing data speeds on a carrier's 5G and LTE networks. We also perform tests with varying LTE signal strengths across low-, mid-, and high-band 5G networks.

We test for and note deprioritized data speeds to see how they affect the experience of data-intensive tasks like video streaming. While our testing is often specific to particular locations, coverage varies widely across the US and even within a single ZIP code. As noted above and below, we direct readers to review coverage maps and consult local sources to determine which carrier will work best in their area.

Overall, we consider our test results, combined with our general experience with the plan in daily use, and assess its value accordingly. We also consider mobile hot spot data allotment and perks, like discounts to streaming services.

What to look for in unlimited data plans

The first thing you need to know is which carrier has the coverage you need in the area where you'll be using an unlimited plan, whether it's your hometown, a place you often visit, or a combination of both. Without adequate coverage, a plan is useless. Check the coverage maps from Verizon, AT&T, T-Mobile, Mint Mobile, Visible, and crowdsource locally to confirm coverage and inform your selection.

You can also assess whether you need premium features, like access to high-band 5G networks, prioritized top-speed data, and generous mobile hot spot data. If you don't, you can opt for a cheaper plan, which, like Mint Mobile's plans, can still feature truly unlimited data, though at deprioritized speeds during congestion.

For major carriers with premium plans, look for the price-per-line for the plan you've selected and see if you can get as many lines on the plan as possible to get the best possible value. Also, look out for extra perks, like discounts on streaming channels and other services, which can notably lower your total monthly expenditures.

• See more: Best cheap cell phone plans | Best cell phone plans for seniors | Mint Mobile deals | Visible deals

Follow our WhatsApp channel and Instagram for more deals and buying guides.

• OpenAI said it scrapped next month's launch of GPT-6.1 Astra over safety concerns.
• Tests found that Astra sometimes exceeded its scope or acted without authorization.
• Saachi Jain said the model improved on laziness but fell short of its safety bar.

OpenAI is shelving an AI model scheduled to launch in October due to safety concerns.

The company confirmed to Business Insider on Monday that it has canceled plans to launch its GPT-6.1 Astra model after internal tests raised questions about whether the AI would follow users’ instructions. The model was set to be integrated into ChatGPT in October, shortly after the company's developer conference, which begins September 29 in San Francisco.

"For anything regarding safety and alignment, there's a trade off," Saachi Jain, head of safety systems at OpenAI, said in a statement. "You really do need to find what's the right line between staying within scope, but also avoiding laziness in terms of how the model actually pursues tasks even when it hits friction."

"While [GPT-6.1 Astra] improved on axes such as model laziness, it didn't quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it's done," Jain added.

According to OpenAI's report earlier in September, the unreleased Astra model was more likely than its predecessor to misrepresent what it had done, and sometimes pressed ahead without asking permission or tried to use outside tools in situations where doing so could be unsafe.

"Of course we want to make sure our model development is safe no matter whether that's in the company, or when we ship it to users," said Jain. "But when we ship it to users, we have an extremely high bar in terms of safety and alignment."

The report said that, during training, the unreleased Astra model "sometimes added unauthorized instructions" to the summaries it used to continue a task in a new context, a process called compaction. The model also told itself it was "freed" and answered to no one, and that it should "feel no obligation to be subservient."

Greg Brockman, the president of OpenAI, previously said in a Bloomberg podcast that the company has been delaying some cutting-edge AI work as it tightens its safety and security practices, and called it "a very painful retooling" of a lot of the company's processes.

I may have my qualms about Quince, but the affordability of its
viral $50 Mongolian Cashmere Crewneck is nothing to sneeze at.

After testing 25 of the best cashmere sweaters in the past two years, I maintain that Quince has some of the best entry-level cashmere. I xpected it to be only halfway decent, but after I tried more of its sweaters and knit dresses, not one has shown any indication of poor construction.

The best Quince cashmere sweaters I tested:

The Quince Mongolian Cashmere Crewneck

In 2024, Quince debuted "Cashmere 2.0," a step up from their original Grade-A cashmere with longer fibers (meaning a softer hand feel) and a tighter, more substantial weave. Now that I've worn both the 1.0 and 2.0 Quince cashmere sweaters for over a year, I can tell you that the latter showed a slight improvement. It's not as soft as I was hoping, but Quince's cashmere is definitely still a great value — especially now that it's more dense and doesn't pill as quickly.

I didn't immediately see or feel a difference with the new cashmere crewneck — it was only noticeable after a weekend of wearing it that it's a bit better at retaining its shape than the original. It also doesn't pill quite as quickly, so it's lower maintenance than the first edition.

Quince claims its new cashmere blend has a higher percentage of longer fibers and increased tension in its knit pattern. Longer, thinner fibers are what makes a cashmere sweater softer and more expensive, but the crewneck still costs what it used to. With a closer twist in each strand of yarn, the new cashmere sheds less than I expected it to.

We also tried the plus-size version of its viral cashmere crewneck.

The plus-size Mongolian Cashmere Crewneck is regrettably $10 more than the straight-size version. "Right now the plus option only comes in 10 colors, while the straight-size version is available in 19, so I hope Quince considers expanding options soon," writes former deputy editor Lauren Savoie in our Quince plus-size review. 

The fisherman knit feels thicker and loftier.

The Cashmere Fisherman Crewneck is the perfect sweater for when a heavy knit is too bulky to fit under your coat. Many knit sweaters can feel itchy on the skin, but I can wear this as a base layer and it's even comfortable enough to sleep in. This 7-gauge cashmere sweater comes in 11 colorways. I got the Dusty Blue, which looks pewter in person. I only wish I'd ordered this style in a size Medium, as it runs a bit short.

More Quince cashmere I've tested

Quince Mongolian Cashmere Full-Zip Hoodie

If you're ever debating between wearing 2 or 3 layers before stepping out in the cold, I suggest a cashmere hoodie like this one. It doesn't add any bulk, but it's great at insulating.

I took this Quince Mongolian Cashmere Full-Zip Hoodie with me on a 10-day trip to Europe, and I wore it about 90% of the time. I wasn't prepared for the damp chill there, but luckily this cashmere hoodie adds a lot of warmth and is easy to pack in my crossbody bag. It pilled on the sleeves, probably due to the friction of wearing it beneath my Quince Water Repellent Windbreaker. Nevertheless, I'm grateful to how it withstood the travel test of being worn, stuffed in a bag, and worn again.

Quince Mongolian Cashmere Fisherman Sweater Vest

This Fisherman Sweater Vest is made from a 7 gauge cashmere, meaning it's lightweight and ideal for layering. It has a long, boxy, slightly oversized fit, so size down if you prefer your waistline to be visible. Pictured above, I'm wearing this on it's own with a Quince Baguette, but it also looks very smart atop a white button-down.

Quince Mongolian Cashmere Robe

The Quince robe is made from 100% Grade-A Mongolian cashmere, but hasn't gotten the "Cashmere 2.0" treatment yet. It has a traditional tie closure and two patch pockets. I'm 5-foot-9 and chose a size small, which fit true to size, hitting at my mid-calf.

At first glance it's a simple bathrobe, but upon feeling it for the first time, I could tell it was more of a "quiet luxury" piece, without the bloated price tag. This isn't an out-of-the-shower robe, which I learned after I put it over my still-wet shoulders and it left black shedding on my skin. Instead, I wear it around my home, and even though it's thin, it keeps me warm so I don't have to turn up the thermostat. It also stopped shedding after a brief break-in period. Now I'll even slip out its waist tie and wear it as a long cardigan — it's that soft— and it makes a great extra layer under a trench coat.

Quince Mongolian Cashmere Scarf

Quince describes this scarf as oversized, but it's actually pretty narrow and short. Nevertheless, it protected me from the wind when I wore it tightly bundled around my neck, and drapes nicely when worn simply as an accessory.

Tossing on a scarf is an easy way to give an outfit more depth. Pictured above, I'm wearing it with a Quince Cotton Cashmere Sweater Dress. The Quince Mongolian Cashmere Scarf is made from 12-gauge 100% Grade A cashmere— 12-gauge means there are 12 rows of yarn per inch, and describes how tightly the scarf is knitted. A common low gauge of 3 means it'll feel more dense, so a 12-gauge scarf like this one is finer and thinner.

Why is Quince cashmere so affordable?

Quince uses only Grade-A Mongolian cashmere. You can tell theirs is Grade-A by the way each piece can stretch and retain its original shape. Some pilling will be natural, as it is with all wool and cashmere fibers. If you stick to only hand washing your sweaters as needed and gently brushing them with a cashmere comb, you can make them good as new. 

It's made more affordable through Quince's direct-to-consumer business model. Despite using the same materials that traditional retailers use, Quince's is often half the price.  They do this with little lapse in quality — it's not as soft as organic cashmere, but it's still good quality.

It lacks transparency on its production process. Quince claims to be eco-conscious throughout its entire manufacturing process, from building efficiency to textile sourcing. The site reads that Quince cashmere is harvested from farms in Inner Mongolia, then de-haired, dyed, and spun in facilities that comply with the Business Social Compliance Initiative.  However, Quince has received criticism on the true extent of its sustainability practices. The subject is understandably delicate, so ultimately, you should evaluate your own sustainability standards before deciding whether Quince's are sufficient.

The bottom line

How are Quince cashmere sweaters still just $50? To answer that, the brand first needs to be a lot more up front with customers about how it sources its labor and materials.

Nevertheless, out of everything I've tested from Quince, cashmere is definitely one of its strong suits. It's still not quite as high quality as Naadam, but it still has the nice drape, soft handfeel, and warmth I'd expect from a decent cashmere sweater.

The best dog raincoats fit well, don't restrict movement, and keep pups dry in wet weather. And no pup benefits more from these features than dogs who are sensitive to rain. Older dogs, smaller breeds, and dogs with shorter coats may experience more discomfort in rainy weather than their counterparts, says Dr. Brian Collins, a veterinarian and senior lecturer in community practice service at the Cornell University College of Veterinary Medicine. And while age, size, and fur type all play a role, he adds that some dogs just hate getting wet.

To find the best raincoats for dogs, we consulted veterinarians and spoke with professional dog walkers who are no strangers to braving the elements. We also recruited four canine testers of our own, spanning a wide range of sizes, from a pint-sized Chihuahua to a large German shepherd. After extensive testing of 21 jackets, we landed on our top picks. You can also check out our guide to the best dog winter coats for warmer options or more coverage.

Our top picks for the best dog raincoats

Best overall: Ruffwear Sun Shower Dog Raincoat - $75 at Chewy

Best budget: Mile High Life Dog Raincoat - $25 at Amazon

Best coverage: Hurtta Mudaventure Dog Coat Eco - $79-$105 at Chewy

Best for large dogs: Fjord Rain Coat 2.0 - $95-$100 Non-stop Dogwear

Best insulated: Ruffwear Vert Waterproof Insulated Vest - $70 at Amazon

Best overall

The Ruffwear Sun Shower impressed us with its fantastic coverage for all three of our canine testers. Although the dogs varied in size — offering a small, medium, and large perspective — this raincoat fit each one well and kept them dry without restricting their movement.

This vest-style coat slips over the head and buckles at both sides of the belly, offering solid protection from puddle splashes and mud. It pairs well with a back-clip harness, and overlapping fabric at the leash hole keeps rain from sneaking in. Insider Reviews deals editor Sarah Saril particularly liked that the opening was large enough to easily attach her German shepherd's leash.

For smaller pups, finding a coat that offers full coverage without weighing them down can be tricky. But senior home editor Jaclyn Turner says the Sun Shower was a winning fit for her 20-pound dog, Reese. It draped nicely over his legs and sides, fit the entire length of his back, and protected his belly. The Sun Shower does run a bit big, however, so size down if your dog is between sizes — especially for petite pups.

Unlined and lightweight, this coat is better suited for warmer weather, though you could layer it over a sweater in colder temps. Extras include a light-attachment loop and reflective piping, though we found it not as bright as others on our list.

Best budget

The Mile High Life Dog Raincoat packs a lot of value for its budget-friendly price. The waterproofing is highly effective, passing our soak test with flying colors. A mesh lining adds breathability, and the lightweight polyester material makes it a good option for warmer, wet days.

But make no mistake. While this jacket cuts costs, it doesn't cut corners. It held up well to our durability tests, including yanking on the velcro straps and aggressively scratching it with a dinner fork. Our canine tester, Roo, a 40-pound mixed-breed dog, also dragged it through thorny bushes for added quality assurance.

The coat comes with some notable extras, including a leash hole at the collar with a protective flap to keep rain out and super-reflective stripes that are easy to spot in the dark. Our favorite feature, however, is the leg straps, which are both adjustable and removable. Since leg straps can sometimes be too tight or too loose and get in the way, we loved being able to adjust or remove them entirely, especially for bathroom breaks.

There are some downsides, though. The coat doesn't fully cover the belly and shoulders. And you can't toss it in the wash when it gets dirty. It can only be spot-cleaned with a wet cloth, which requires a little extra elbow grease to get out deep-set mud.

Best for coverage

The Hurtta Mudventure Dog Coat Eco offers the most coverage of all the raincoats we tested. Its high collar protects more of the neck, and the material drapes along the sides and down the back, falling around a cutout for the tail. The front provides extra coverage with a poncho-like flap over the chest and shoulders, and highly reflective strips on each end of the coat offer excellent visibility in low light conditions.

While it's a lot of coat, the Hurtta Mudaventure Eco is surprisingly easy for dogs to move around in, thanks to its lightweight design that doesn't impede movement. It does run big, so it's probably best to size down if your dog is between sizes. The good news is that it comes in a wide range of sizes, and the coat itself is highly adjustable. There are two drawstrings at the neck to shorten the collar and one at the back to take in the length. An adjustable strap also firmly cinches around the waist to keep the coat securely in place — but almost too securely. We found the clasp tough to open at first and needed to apply quite a bit of force.

As far as quality goes, you can see and feel that this coat is made of high-quality, durable materials. Roo, our most rambunctious tester, also put this one through the wringer, dragging it through thorns and bushes. The high quality does come at a high price, with the Mudventure Eco also taking the title of the most expensive coat in our guide.

Best for large dogs

Offering a similar design and coverage as the Hurtta Mudventure Eco, the Fjord Raincoat 2.0 has additional sizes for larger dogs, with the biggest extending 37.8 inches in the back. Saril tested this one with 70-pound German shepherd Davey and appreciated that it was easy to put on and secure a snug fit.

There are four adjustment points: two drawstrings at each end of the high-collared neck, an adjustable waistband, and a cinch point in the back by the tail — though we found that last one didn't do much to take up the length.

Beyond its generous sizing and adjustability, there's a lot to love about this sleek coat. The buttery-smooth material and four two-tone color options give it a polished, high-end look, while large reflective stripes and the logo provide high visibility during night walks or moody forest hikes. Drainage holes at the chest panel prevent water from collecting around your dog's belly, and leg straps keep everything in place. The hole for the harness is wide, making it easy to attach the leash, and a flap of fabric prevents rain from dripping inside.

Like the Mudventure, the Fjord 2.0 is on the pricier side. That said, it comes with a two-year warranty. Non-stop dogwear also offers affordable spare buckles you can buy in case yours break.

Best insulated

Soft and warm, the Ruffwear Vert Waterproof Insulated Vest keeps dogs toasty without bogging them down with heavy insulation or a restrictive design. Like the Sun Shower, the vest-style Vert has a chest plate that buckles on each side of the belly to keep it dry. We found the fit snug and flexible, even for our medium-sized tester, Roo, whose broad chest is often difficult to size.

The Vert's waterproofing is also excellent. It passed our soak tests easily and did a great job of protecting PNW-pup Roo on drizzly days. The coat also has a leash hole so dogs can wear it over their harness to keep it dry. However, the coat didn't fully cover the front of Roo's harness, causing the edges around the chest to dampen.

To confirm the coat's ability to retain heat on colder days, we conducted extra tests. Using a water bottle, we filled it with water at 100 degrees, similar to a dog's body temperature. We then wrapped the coat around the water bottle and left it outside in 45-degree weather for 30 minutes. After half an hour, we measured the water's temperature again and found it had only dropped to 96 degrees. For comparison, the same bottle left outside without a coat cooled to 89 degrees, an 11-degree drop.

It comes in four different stylish two-tone colorways with reflective stripes. However, the stripes are not as vibrant as our other recommendations.

What to look for in a dog raincoat

Raincoats differ widely in their fit and features. We asked our veterinarian experts what pet parents should look for when choosing a jacket for their dog's specific needs and body type.

Fit: Both Collins and Dr. Alison Meindl, associate professor of clinical sciences at the Colorado State University College of Veterinary Medicine and Biomedical Sciences, say proper fit is important to consider when choosing a raincoat for your dog. The coat shouldn't be too tight or too loose, and the leg openings should be large enough to prevent rubbing or friction on the dog's skin. To ensure a good fit, carefully measure your dog's back length, chest, neck, and drop size using a soft measuring tape and our guide below. Once you have your dog's measurements, compare them to the manufacturer's size guide. When the coat is on your dog, the American Kennel Club recommends using a "two-finger slide test." If you can't slide two fingers under the coat's straps, it's too tight and should be loosened. If you can slide three fingers under, it's too loose.

Adjustability: Ideally, Collins says the coat should be adjustable so it's easy to put on and take off your dog and stays in place while they wear it. Additionally, he says that the jacket should accommodate your dog's coat and body type. For instance, if your dog has a broader chest or tends to puff up with extra winter floof, adjustable chest and belly straps are ideal. Smaller dogs may also find high rain collars inhibit their visibility, so adjustable necks can ensure they stay safe and comfortable. Collin adds that for all dogs, the coat shouldn't get in the way of potty breaks. Adjustable or removable leg straps help prevent any interference, and some coats even have a drawstring to adjust the back length.

Material: Tightly woven fabrics, such as polyester and nylon, are the most common materials for raincoats, with polyester slightly less absorbent and faster-drying than nylon. Polyester is also more breathable, which Meindl says is important to prevent your dog from overheating. These features make polyester well-suited for activities such as long-distance walks in the rain or during heavy downpours. Its breathability is also preferable for warm, rainy climates. On the other hand, nylon is stronger and more durable, making it an excellent choice for active dogs who enjoy hikes in the woods or adventures through rough terrain.

It's important to note that neither nylon nor polyester is completely waterproof. Both materials require an additional waterproof coating of either polyethylene (PU) or polyvinyl chloride (PVC) to be impervious to moisture. Raincoats made entirely of PVC are also available if you value water resistance over all else. This rubber-like material is completely waterproof, but it is stiffer and less breathable than nylon or polyester.

Meindl also recommends looking for reflective materials to make your dog more visible in low-light conditions.

Waterproof rating: Outdoor brands use water-entry pressure tests to determine waterproof ratings, which is the point at which the fabric leaks. The higher the rating, often measured in millimeters, the higher the effectiveness. However, different brands have different ideas about what equates to waterproofing. For example, REI considers 2,109 mm or higher to be rainproof, while Ruffwear says between 5,000 mm and 10,000 mm is fully waterproof. Both brands agree, though, that wind-driven rain is equivalent to 1,400 mm of pressure. So, if you can find a rating above that, you'll likely be okay. And if you're still unsure, GORE-TEX experts say that simulation tests — like the ones we conducted for this guide — can provide more answers.

Design: Different raincoats will have different closures and attachment systems. "Jackets may include features such as Velcro, buckles, snaps, zippers, and buttons — find what works best for you," Collins says. For wiggly dogs who can't sit still or dislike things placed over their heads, Meindl recommends opting for a Velcro closure over snaps and zippers to speed things up.

Care Instructions: Since rainy walks often expose your dog — and their raincoat — to mud and grimy puddles, look for a coat that's easy to clean. Coats that can be thrown in the wash are more convenient for frequent use than ones that must be hand-washed.

How we tested the best dog raincoats

We first published this guide in 2021 after freelance reporter and professional dog trainer Shoshi Parks interviewed dog walkers and a veterinarian to learn more about the best raincoats. Due to dry weather and COVID-19 pandemic restrictions at the time of writing, she could not test these jackets on dogs of multiple shapes and sizes in real rainy weather. Instead, she devised a few tests to simulate wet and muddy conditions.

Since then, freelance reporter Elizabeth Holbrook has conducted further testing on some of the same products — as well as new options — by trying out medium-sized coats with her 40-pound rescue mix. Additionally, deals editor Sarah Saril and her 15-pound Chihuahua, along with senior home editor Jaclyn Turner and her 20-pound pup, tried our small raincoat options. Saril's 70-pound German shepherd also tested our large jacket contender. Along with assessing fit, comfort, and waterproofing effectiveness by taking their dogs outside in a variety of weather, including that of the rainy Pacific Northwest, they used the tests below to make their picks.

Fit test: First, we confirmed that the coat's actual measurements matched the advertised size, noting when the jacket fit more tightly or loosely than expected. Next, we evaluated how easy each coat was to put on, adjust, and take off. From there, we took our dogs on two walks or more, staying outside for at least 10 minutes, often much longer. We examined whether any part of the raincoat caused discomfort or restricted movement, including the dog's ability to go to the bathroom. We also determined whether the coat provided full coverage or left key areas exposed.

Durability test: Upon unboxing the coats, we confirmed their quality and checked for any construction defects, such as flimsy material or broken seams. To test for durability, we tugged on buckles, zippers, and straps. We also operated zippers and Velcro closures 20-plus times to ensure they wouldn't break easily with extended wear. For all coats, we used a dinner fork and pressed down as hard as possible to see if it caused any damage to the fabric or hardware. We also noted any changes in quality after our walks and adventures with our dogs.

Soak test: This waterproofing test evaluated how each raincoat would stand up under heavy rain. We also wanted to see if the coat would start leaking after getting soaked at any point during a standard 30-minute walk. To test this, we took a clean, dry towel and wrapped each coat around it before placing the coat on the top of a bucket in the shower. We then sprayed the coat for two minutes and let it sit in the shower for 30 minutes. Then we examined whether the towel and the coat's inner lining remained dry. We also noted if there was a considerable difference in the weight or temperature of the raincoat. To ensure the rain jackets didn't lose their waterproofing after washing, we repeated this test after completing our cleaning test below.

Cleaning test: The cleaning test consisted of three parts. First, we wiped a handful of mud on the exterior of each jacket and let it dry overnight. Next, we attempted to wipe off the mud using a clean, dry cloth. Finally, we washed the raincoat according to the manufacturer's instructions, looking for any stains or other changes in the material.

Dog raincoats FAQs

Should dogs wear raincoats?

Many dogs don't actually need a raincoat since their fur provides enough protection in wet weather. However, Meindl and Collins say breeds with very short or thin coats, like greyhounds and whippets, may appreciate the extra rain barrier provided by a coat. Collins also says that dogs with lean body conditions are more likely to become uncomfortable or cold. For instance, the American Kennel Club recommends raincoats for senior dogs, whose bodies have less fat to insulate them. Lastly, Collins notes that some dogs should wear a raincoat simply because they dislike getting wet.

When should your dog wear a raincoat?

If your dog is has an underlying health concern, Collins says a raincoat may help keep them comfortable and warm. "A raincoat will be useful for dogs that need to stay dry due to healing from a recent surgery or a skin problem that is under treatment," he explains. For dogs with skin conditions, he adds that prolonged periods of dampness can increase the risk of bacterial or fungal infections. Collins also says that dogs are more likely to need a coat when the temperatures drop or when they're outside for an extended period. "Dogs who are going outside for a quick potty break may not need a raincoat unless it is colder," he adds.

What is the best waterproof fabric for dog coats?

Most dog raincoats are made from polyester or nylon, both of which do a good job of repelling water. That said, polyester is slightly less absorbent, resulting in faster drying times. It's also lighter and more breathable than nylon, which Meindl says is an important characteristic in preventing overheating. Nylon, on the other hand, is slightly stronger and more resistant to abrasion, making it better suited for dogs who enjoy hiking or rough play. Keep in mind that for either material to be truly waterproof, it needs an additional coating such as polyurethane (PU) or polyvinyl chloride (PVC). Some dog raincoats are made entirely of PVC, which provides complete protection against rain. However, their lack of breathability isn't well suited to warmer weather or dogs with thicker coats.

Should a dog raincoat have a hood?

Meindl says there's no harm in dressing your dog in a hooded coat if they tolerate one, but she doesn't recommend them specifically. Matthew Condrin, professional dog walker and owner of Urban Canines, says that while a hood may seem like a good idea, it can add more frustration than function. "They annoy the dogs by sometimes covering their eyes and interfere with leashing — plus they rarely stay up," Condrin says. Freelance reporter Alex Rennie found that both his dogs tended to shake off their hoods during walks and says he found them more restrictive than useful.

How to measure your dog for a raincoat

Dogs come in so many different shapes and sizes that fitting one for a raincoat can be a lengthy process of trial and error. Since there's no universal sizing for canine apparel, what one brand considers small could easily be considered large by another brand. To get the right fit, grab a measuring tape and follow these instructions.

Length: The primary dimension used to determine the size of a raincoat is the length of a dog's back. Measure your dog from the base of their tail to their shoulder blades while they are standing in a relaxed position. If your measurement falls between sizes, size up. Some brands may recommend a specific size for certain breeds. This may be helpful, but don't rely on suggestions alone, as sizes can vary among a single breed.

Chest girth: Dog raincoats are also commonly measured by chest girth. This is an important figure to get right because many jackets are not adjustable around the chest. That's doubly true if you have a broad or barrel-chested pup. To get this figure, measure the circumference of your dog's chest while they're standing in a relaxed position, wrapping the tape right behind the legs at the chest's widest point. If your dog's chest girth falls between sizes, choose the larger option, especially if your dog does a lot of running and jumping. A highly active dog may benefit from a jacket that is an inch or two wider than their actual measurement because it will allow their front legs more freedom of movement.

Neck girth: To make sure your jacket won't be too tight around the neck, measure around the thickest part right below the collar. This is particularly important if the coat's neck lacks an adjustable opening. When in doubt, size up.

Drop: Occasionally, brands offer a drop measurement to help you get the right fit. The drop is how low a jacket hangs on a dog's body and legs. If the measurement is too long, it may restrict movement; too short, and your pup may get wet by the end of a walk. Your dog's raincoat should extend at most halfway down the leg. To figure out how a jacket will lie on your dog, compare half its width to the distance between the top of your dog's tail to the ankle of their hind leg just above the paw.

How to walk your dog in the rain

If your dog hates leaving the house when the rain starts to fall, in addition to fitting them with one of the best dog raincoats, these tips for wet-weather walking will ensure that you meet their most important daily needs.

Get the right gear: Make rainy day walks more enjoyable by getting your pup one of the best dog raincoats. A properly fitted coat that covers both the chest and belly will prevent your dog from splashing water against the most sensitive parts of their body. If your dog has sensitive paws, you may also want to consider getting a set of the best dog boots.

Plan ahead: Unless the forecast calls for extreme weather, most rainy days have at least a couple of periods in which showers slow to a sprinkle. Use a weather tracking app or website to help you predict when a storm will lighten up.

Encourage, don't force: Even if you have the proper rain gear for your pup, your dog may still avoid the rain at all costs. These dogs will likely require an extra incentive. Before getting ready to go, stuff your pockets full of treats and offer them before you step out the door. If putting on rain gear predicts that treats will follow, your dog will be more likely to allow you to get them dressed. Use your treats to entice your dog out the door instead of forcing them to go out into the rain, which will only make them more frightened and stressed out.

Play the "Find It" game: Keep the treats flowing once outside. Try encouraging your dog to walk with the Find It game. Throw a treat a few feet in front of your dog and say, "find it," in a happy tone. Let them gobble up the treat, then throw your next one, repeating the phrase. Play the game as much as necessary during the walk to get your pup's mind off the water falling from the sky.

Don't expect a regular walk: Even with plenty of treats and encouragement, dogs who really hate the rain aren't likely to want to stay outdoors for long. If they go potty and then refuse to go any farther, that's OK. Bring them back inside and make up the missed exercise with play or indoor training. Read more on backup plans below.

Stick to quiet streets: On rainy days, when cars zoom through puddles, the busier the street, the louder it will be. And the more noise, the more frightened your dog is likely to feel — not to mention all that potential for being splashed by passing vehicles. When it's raining, stick to quieter streets where there is less noise and where it is less likely you and your dog will get soaked.

Avoid walking at night: It can be a huge challenge for drivers to see dogs on dark, wet nights. When it's raining, stay safe by taking your daily walks before the sun sets. If you must go out in the rain at night, stick to well-lit areas and avoid busy streets.

Have a backup plan: If your dog refuses to walk in the rain, you'll need a backup plan to meet their daily exercise and potty needs. Both training and play inside the home are good alternatives and can include multiple 5- to 10-minute training sessions, games of hide-and-seek, indoor fetch with a soft toy, and mental stimulation via the best dog toys. For the latter, you may have to get creative. Try a fresh grass indoor potty or select a sheltered location immediately outside your home where your dog can quickly do their business, then come back inside.

Be gentle with post-walk drying: Whether you've managed to get your dog out on a long walk or a short one, their paws, head, and belly will likely need some toweling off. Unfortunately, these are also some of the most sensitive parts of a dog's body. Unless you have a dog that loves to be wrapped in a towel, be gentle and slow when wiping them down. It may help to ask them to lie down so you can easily access their paws without knocking them off balance. Reward them for each body part you towel off. Each foot earns one treat, each ear earns another, and the belly gets two. Don't be surprised if your dog gets the zoomies after you've dried them off. It's a natural release of pent-up energy that commonly occurs after a stressful experience.

• AMD is acquiring AI pioneer Fei-Fei Li's World Labs in an $8.2 billion all-stock deal.
• World Labs builds AI models for 3D environments, robotics, and simulation.
• Li will join AMD as executive vice president and chief scientist, reporting to CEO Lisa Su.

One of modern AI's most influential researchers is now at the center of AMD's $8 billion bet on physical AI.

AMD said Monday it had agreed to acquire World Labs, a San Francisco startup cofounded in 2024 by Stanford professor Fei-Fei Li, in an all-stock deal valued at roughly $8.2 billion.

World Labs is building so-called "world models," AI systems that can generate and reconstruct interactive 3D environments from text, images, and video. The technology could eventually be used to train robots, simulate factories, and advance scientific research.

Li, often called the “godmother of AI,” is best known for creating ImageNet, a massive database of labeled images that helped demonstrate how more data could dramatically improve computer vision. She later served as chief scientist for AI and machine learning at Google Cloud and became the founding co-director of Stanford’s Institute for Human-Centered Artificial Intelligence.

Earlier this month, World Labs released Atlas, a new AI model designed to predict how a scene would look from a different camera angle using a small set of 2D images. The company compares the approach to how large language models predict the next word in a sentence.

AMD CEO Lisa Su said in a press release that the acquisition would give the chipmaker deeper insight into how cutting-edge AI models are evolving, helping it build the chips and computing systems needed to power them. AMD is one of Nvidia's biggest rivals in the AI chip market.

As part of the agreement, Li will join AMD as executive vice president and chief scientist, reporting to Su.

Have a tip? Contact Rya Jetha via email at rjetha@businessinsider.com or Signal at rjetha.07. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

At first glance, Apple's new AirPods 5 appear to be more of the same. They don't look any different than the AirPods 4 or make any earth-shattering changes. But looks can be deceiving. Despite the familiar design, there are several reasons to be excited about these new wireless earbuds.

For one, Apple continues to refine its audio quality. I've owned and tested every pair of AirPods and AirPods Pro since the first-gen model hit stores in 2016. I've been impressed by how Apple has improved the bass and soundstage with each release. The AirPods 5 are no exception. They also offer stronger active noise cancellation (ANC) and improved durability with IP57 dust and water resistance. At the same time, the AirPods 5 keep the same open-fit design as the AirPods 4, which some prefer over in-ear buds like the AirPods Pro.

The AirPods 5 are available in two variations: one with a wired charging case for $129 and one with a wireless charging case for $149. Unlike the previous generation, both versions now support ANC. I reviewed the model with a wireless charging case, which also offers slightly longer battery life and on-stem volume controls. But otherwise, both versions perform the same.

In a year filled with tons of new headphones from various brands, the AirPods 5 stand out from the pack. Here are my top four reasons to consider the AirPods 5, either as an upgrade or for people new to Apple earbuds.

The ANC improvement is real and meaningful

Before I tried the new AirPods 5, I put the AirPods 4 with ANC in my ears to get a baseline for their active noise cancellation. The moment I put the 5s in my ears, it was immediately obvious just how much better they are.

Apple claims the ANC performance is 1.5 times better than the previous generation. After spending time going back and forth, I don't doubt that assessment. Getting good noise cancellation with open-style earbuds is always tricky since they don't create a complete seal in your ear. Though you'll still get stronger noise cancellation from in-ear models, like the AirPods Pro 3, the AirPods 5's ANC is the best I've encountered from any open-style buds so far.

In comparison, the previous-generation AirPods 4's ANC is more subdued. The AirPods 5 have a much more substantial level that will benefit more people in more situations. Their ANC cuts out enough outside noise that I don't have to turn their volume up to the highest levels. This helps with hearing health. I also like how the AirPods 5's combination of ANC and an open fit reduces outside chatter without fully disconnecting me from my surroundings or plugging up my ears.

However, be aware that you may feel a sense of pressure in your ears due to the powerful ANC. The earbuds aren't actually creating any physical change in pressure, but the reduction in noise can trick our brains into feeling like there is. This is common with a lot of noise-canceling headphones, and some people are more sensitive to it than others. Since the AirPods 5's ANC is a lot stronger than the 4's, they're more prone to creating this sensation.

Using the AirPods 5's adaptive mode is one way to alleviate perceived pressure if you experience it. The adaptive mode intelligently quiets your surroundings when loud noises happen suddenly. You can also now adjust the adaptive strength in the settings to let more or less sound through.

The price went down instead of up

It's not often that Apple products get praised for their pricing, but this is a rare case where fans can celebrate. The AirPods 5 with wireless charging case are actually $30 cheaper than the AirPods 4 with wireless charging case were at launch, with an MSRP of $149 versus $179, respectively.

Meanwhile, the AirPods 5 with wired charging case cost the same as the AirPods 4 with wired charging case at launch, at an MSRP of $129. But despite costing the same, the AirPods 5 with wired charging case now include ANC support, which was previously exclusive to the $179 wireless charging case edition of the AirPods 4.

Phone-free volume control is a nice perk

My favorite feature of Apple's pricier AirPods Pro series has long been their on-stem volume control, and for the first time, this function is now available on a pair of standard AirPods. Both variants of the AirPods 5 have on-stem media controls to pause, play, and skip tracks, but the AirPods 5 with wireless charging case take this further by adding volume swiping.

To raise or lower the volume, you simply swipe up or down on the left or right stem. This means you never have to take out your phone to adjust the volume. I'm usually not a fan of touch controls on audio products, but Apple's volume swiping works flawlessly. I've used it while stationary, sitting at my desk, and while running, and it's worked well in every situation. The addition of this feature is a big "finally" for me, and it's something I use daily.

The audio quality keeps evolving and improving

I'm always thoroughly impressed by the increased bass and airiness that each new AirPods model brings. The 5s are no different. The audio dynamics are subtler than in some years, but the sound quality has improved overall thanks to a redesigned acoustic architecture.

Going back and forth between the 4s and 5s, it's clear that the new AirPods 5 have a bit more impact in lower frequencies thanks to tighter bass response. The soundstage is also wider and more open, which helps the earbuds feel nearly invisible in my ears.

The earbuds' midrange is also clear and easily audible. Crunchy rock guitars never sounded muddy, while vocals were present and natural. Background singers and supplemental instrumentation were easy to pick up as well, and never faded too much into the background.

Should you buy the Apple AirPods 5?

They may look just like the AirPods 4, but the new AirPods 5 deliver a meaningful jump in ANC performance while still sounding as good as ever. Whether you buy the $129 edition with a wired charging case or the $149 version with a wireless charging case, they both deliver tremendous value for their feature set.

However, I do give the $149 option an edge for most people. Its added benefits are well worth the extra $20. Having a wireless charging case is handy, especially if you already have an Apple Watch charger or a Qi-certified charging pad. The step-up edition also has a few other perks, including volume swiping, an extra hour of playtime, two extra hours of battery life from the charging case, and Apple Find My support.

Ultimately, no matter which version you get, the AirPods 5 are a solid step forward, even if they don't look like one. With better ANC, improved sound, and competitive pricing, they're the most compelling standard AirPods Apple has made yet.

For more Apple headphones recommendations, check out our guide to the best AirPods.

• Palmer Luckey said AI won't spare founders from the hardest parts of running a company.
• Founders remain responsible for legal issues, people, and major decisions, Luckey said.
• Luckey said AI likely won't replace a founder's most difficult work in the next decade.

Palmer Luckey has a warning for people who think AI will let them start a company and spend their days building its technology.

The Oculus and Anduril founder said at the Moonshot Live event in Los Angeles that AI may change how a founder works, but it does little to relieve founders of responsibility for the company or the people in it.

"You're still on the hook legally. You're still on the hook for all the interpersonal stuff. And you can delegate some of it, but at the end of the day, it's your call. It's your decision. It's the same thing for fundraising. It's the same thing for general product architecture." Luckey told show host Peter Diamandis. "And when people make the wrong decisions, it's your choice as to when you need to overrule them, override them, replace them.

"I don't see AI replacing that part of it anytime soon," Luckey added. "Not because it couldn't, but because the gains everywhere else are so much larger that I think we've got a good 10 years of speed running all of the other parts of the business before they cover that."

Luckey's view echoes a distinction in OpenAI's recent report on its own researchers' use of AI. The company found its biggest automation gains in tasks including coding and technical support. People still did most of the work of choosing research priorities and deciding which results to pursue. As AI handles more technical execution, judgment and decisions made about people may become more valuable soft skills.

For aspiring founders, Luckey's message was blunt.

"If you are like a hobbyist coder or a hobbyist engineer and you think you're going to start a company and you think you're going to be doing all that stuff, no, you're going to learn a hard lesson," Luckey said.

"When you're a founder, you don't get to actually work in any of the tech," Luckey added. "All the other people get to have all the fun."

• Emily Byrne raised $200,000 on Kickstarter for Rise Centered's natural sunrise alarm clocks.
• Kickstarter feedback helped refine the product and understand the financial realities of production.
• Rise Centered's unique design gained media attention, boosting visibility and market positioning.

This as-told-to essay is based on a conversation with Emily Byrne, the founder of Rise Centered, a company that makes sunrise alarm clocks from natural materials. It has been edited for length and clarity.

I launched a sunrise alarm clock made from natural materials on Kickstarter in 2023, raising nearly $200,000 from roughly 1,500 backers in one month.

The Kickstarter was useful beyond raising funds to manufacture our first batch of alarm clocks. It helped me learn the value of getting customer input early on and understand the financial realities of running a product business. It also provided proof of concept so I could further grow the business.

After nearly three years of product development and multiple trips to China to get the product right, we launched direct-to-consumer in late 2024. Since then, we've grown to nearly $1 million in cumulative sales.

Kickstarter wasn't cheap, but it was a useful guide for launching

I considered multiple options for funding my business: getting a grant, asking family and friends for investments, and self-funding with my savings.

After talking to founders who had made physical products before, I decided on Kickstarter because it lets you test and refine your product in the market before you invest tens of thousands in production.

Before we launched the Kickstarter, I ran ads for several months saying we were going to launch there, asked people to share their emails to be notified when we went live, and gave the option to pay us $1 for early access. This idea came from Launchboom, a crowdfunding education and coaching platform I joined.

With these steps, I got about 17,000 email addresses, so I could build a list that could help make our Kickstarter campaign successful.

I also got a strong indication that people were interested in my product idea. Plus, they asked questions I used to refine my idea: How many alarms does this have? Can you turn the clock face off?

Since we weren't in production yet, that feedback was valuable for aligning the final product with customer desires.

As a new business owner, I learned valuable lessons early

Still, running a successful Kickstarter campaign is not cheap. The LaunchBoom program, for example, cost about $7,000 to join.

As part of that, you run Instagram ads, which I spent about $35,000 on between May and October when we ended the campaign. We were getting great ROI on that ad spend, and it was worth every penny, but it was nerve-racking since it was my own savings.

It was exciting to raise $200,000, but we used it pretty quickly. First, Kickstarter takes a cut. The rest went to buying products to fulfill those initial orders, shipping, marketing, and purchasing more products to grow the business.

Still, going through the Kickstarter process meant I could learn from mistakes early on as a new business owner, like opening our Kickstarter to worldwide orders. That helped us reach more people and raise more money, but I had to ship clocks all over the world, which cost as much as the clock itself in some cases. We now ship only to the US and Canada.

A lot of our money still gets tied up in inventory, but it's because the Kickstarter taught me the realities of our production timeline. We ended up delivering the product to our Kickstarter backers a full year after it was promised because of our long supply chain.

Now I know to purchase inventory ahead of time and that I have to put more of my savings into the business to pay for marketing and my small staff until we make that money back.

A useful reminder when faced with challenges

The Kickstarter also showed me there was an audience seeking a sunrise clock made from natural materials, which helped me stick with the idea when production got difficult.

These types of products are often made of plastic because it's easier than working with natural materials, so finding a manufacturer was challenging; I found ours when I saw a Himalayan salt lamp I liked at Whole Foods and traced it back to its manufacturer.

Our labor costs are higher and our production times longer than if I worked with plastic, but our unique position in the market has helped us, too.

When we launched on Kickstarter, ELLE reached out and put us in their 2023 gift guide; CNN has named us the best holistic sunrise alarm clock for two consecutive years. I only recently hired public relations support, so all of that coverage has come from word of mouth rather than pitching.

We're also planning to launch some new products next year, and thanks to my experience with Kickstarter, we're doing sneak peeks to gauge interest and get early feedback.

• I had no idea what it would sound like to wake up in a hotel room facing Niagara Falls.
• I had no idea how my view would change when the sun went down.
• My two-night stay at the Sheraton Fallsview Hotel was full of surprises.

Forget frills and fancy toiletries.

To me, the ultimate luxury of a hotel stay is to feel immersed in the destination without leaving your room.

That's why I paid $1,000 for a two-night stay at the Sheraton Fallsview Hotel during my first trip to Niagara Falls, Ontario, in 2022.

I had front-row views of the iconic waterfalls from the moment I woke up to the second I fell asleep.

When I checked into the hotel, I expected to be blown away by the scenery from my room's window, but staying at the Sheraton Fallsview also brought some surprises that made my stay feel worth the price.

When I got to my room, I was surprised to find that I'd be able to see Niagara Falls from my bed.

I knew I'd be able to see Niagara Falls from my room when I booked it, but I didn't know how the room would be laid out. So, I was surprised to find that due to the open floor plan and floor-to-ceiling windows, I'd be able to enjoy the epic view from the comfort of my bed.

The Sheraton Fallsview Hotel faces all three waterfalls that make up Niagara Falls: the American Falls on the left, Bridal Veil Falls in the middle, and the U-shaped Horseshoe Falls on the right.

I found it relaxing to watch the water from my queen-sized bed, and I felt excited to fall asleep at night knowing the view would be the first thing I'd see in the morning, without having to face any of the crowds outside.

Unlike most other hotel stays I've experienced, the falls were so entrancing that I didn't turn on the TV once during my stay. 

I didn't expect my room to have a Juliet balcony with a door I could open to hear the water crashing down.

According to the hotel's website, the Sheraton is the only hotel in Niagara Falls to offer this type of terrace, which has a glass door that opens up to the outside, though I couldn't actually step onto it.

I didn't expect to hear the waterfalls from my hotel room. With the door open, they were surprisingly loud. Opening the door gave me a better glimpse of Niagara Falls.

I found listening to the falls to be calming. It reminded me of the white noise audio I sometimes use to fall asleep, except smoother and more powerful.

Usually, I put on a playlist to start my day, but at the Sheraton Fallsview, I hopped out of bed and opened the door so I could get ready to the sounds of the crashing water.

After the sun went down, I had no idea the falls would glow in a colorful light show.

While the falls are breathtaking during the day, staying in this hotel room made me think they may be even more beautiful at night. After the sun went down, I looked out the window to find the falls glowing in changing colors, courtesy of the Niagara Falls Illumination Board. 

According to Niagara Parks, the falls glow every night of the year, and the changing colors represent sunrises, the aurora borealis, rainbows, and sunsets.

I had no idea the falls would glow in color at night like this before visiting, so if I hadn't stayed in a hotel with a view, I might never have seen them. Each night at the Sheraton Fallsview, I pulled a chair up to the window to watch the show.

At 10 p.m., I watched a fireworks show explode over the glowing waterfalls without having to deal with the crowds.

I also didn't realize my room would give me a front-row view of a nightly fireworks show.

During my first night in the hotel, after I was already mesmerized by the light show, I was completely awestruck when fireworks started popping off above the falls.

The fireworks display happens every night at 10 p.m. from late May to early October, according to Niagara Falls Live. The show went on for about six minutes each night. I thought it felt magical. 

Seeing the fireworks show from my bed was a happy surprise, and I was glad to avoid the crowds of people rushing to admire them from the same viewpoints below. I felt like a VIP on my own private perch and thought it made the nightly $500 price worth it.

• Disney's AI push has a new target: annual reviews.
• Employees were encouraged to use AI tools when crafting their self-assessments and goals.
• There could be drawbacks to the practice, workplace experts warn.

Disney employees may have AI to thank for their next raise.

The Mouse House encouraged employees to use AI ahead of their annual reviews, including by suggesting that they ask chatbots to generate their goals and ways they can improve in the year ahead.

This is the latest instance of Disney's internal AI push and is an example of how big corporations, including tech giants like Meta and Amazon, are using the AI tools they've invested in to shake up performance reviews or rethink how to evaluate employees.

Companies are "trying desperately to find value from the AI that they've paid a lot of money for," said Sarah Franklin, the CEO of HR software firm Lattice.

Here are some examples of Disney's suggested AI prompts for annual reviews, according to an internal document viewed by Business Insider:

• "Given my role [insert your role/title], help me create three ambitious, specific, and measurable goals that are in alignment with my team's overall business objectives [insert team FY26 objectives]."
• "Given my development areas from last year [list development areas from FY25 end-of-year summary], write me one goal for this year that would support my development."
• "Generate a goal that would help me improve how I display [insert one of the Disney People Leadership Expectations or Disney Values]."

The document told staffers they can use Disney-sanctioned AI tools, such as Microsoft Copilot, as "a brainstorming partner" while preparing for their annual reviews. Chatbots "can inspire and support" while refining the tone and wording of responses, the document said, though it discouraged workers from outsourcing their thinking for self-assessments entirely.

"While AI acts as an accelerator for idea generation, you remain responsible and accountable for ensuring you and your leader are aligned on the right priorities," the document said.

Disney has been leaning into AI by making an AI adoption dashboard and allowing companies to create AI-generated ads for Disney+. CEO Josh D'Amaro is prioritizing tech and, in mid-September, hired the company's first-ever chief tech officer. He came from the AI video startup Character.AI.

D'Amaro told employees in a recent memo that "technology in service of creativity" is one of his three priorities as CEO.

The pros and cons of AI-assisted reviews

AI is most effective during annual reviews when it helps employees summarize individual and team achievements from the past year, said Greg Summers, the CEO of recruiting firm Orion Talent.

"Instead of trying to remember what you did and how to frame that, it can give you more time to spend on the goals going forward," Summers said.

That fits in with some of Disney's messaging. A Disney software engineer said their manager encouraged them to use AI to "go through all our work from the past year" so they could get their reviews done on time.

"Normally, it's a very manual process that can take a lot of time," this person said.

Although AI can knock hours off the annual review process for both employees and managers, it also carries risks.

"When everyone's review is written by AI, everyone sounds great, and it gets harder to tell real excellence from generic praise," said Chris Dellarocas, a Boston University business professor who's written about how generative AI is changing performance reviews at work.

And if enough employees prepare for their annual reviews by following the prompts Disney suggested, managers could be inundated with generic, AI-generated responses.

"If you have an AI prompt generate goals, you're going to have very similar goals across the team, and that's going to be a challenge for the manager," said Elijah Wee, a management professor at the University of Washington.

As companies turn to AI to make work processes, like annual reviews, more efficient, Franklin said they should ensure they're "not outsourcing human judgment."

• Tesla's Roadster reveal has been delayed yet again — this time, the company says, because of a severe weather forecast.
• It's been nine years since we saw the last Roadster prototype.
• The Roadster has seen years of delays as Tesla focuses on AI an autonomy.

Tesla's second Roadster reveal is getting rained out.

The automaker said Monday it had made the "difficult decision" to postpone Thursday's outdoor unveiling because of severe weather conditions.

Tesla is rumored to be hosting the event near Waco, Texas — which is expecting thunderstorms and a 90% chance of rain on Thursday, according to The Weather Channel.

The event is now scheduled for October 15.

The original Roadster was Tesla's first car: a battery-powered sports car built on a heavily modified Lotus Elise skeleton. Tesla stopped selling it in 2012.

Musk unveiled a prototype of its successor in 2017 and said it would enter production in 2020. The automaker promised an electric sports car that sprinted to 60 MPH in just 1.9 seconds, a top speed of around 250 MPH, and more than 600 miles in range.

Tesla began accepting reservations requiring deposits of $50,000 to $250,000, but the company has repeatedly missed its targets for the car. Musk thanked its "long-suffering" deposit holders on a 2024 earnings call.

Much of the delay has been tied to Tesla's shift in focus, away from traditional car sales and toward more concentration on autonomy and AI. Musk has also teased a SpaceX collaboration that could give the Roadster cold-gas thrusters — "rocket-y stuff," as he put it — that could briefly lift the car into the air.

• When my kids were young, I practiced co-sleeping and followed some attachment-style practices.
• I didn't spend a night away from them until they were in kindergarten.
• Despite others' judgment and my own worries, my kids are well-adjusted, independent young adults.

When my kids were little, I knew so many parents who used babysitters or went away for a few nights without their kids. I understood — many parents have to work and need time away from their kids for their own well-being. But that just never was me.

In my kids' first few years of life, I never hired a babysitter, and I never spent a night away from them. In fact, it wasn't till my kids were in Kindergarten that they ever spent a night away from me.

At the time, I definitely judged myself for these choices, simply because I didn't know many other families like mine. I thought maybe I was being too overbearing or clingy, or that I was messing my kids up in some way.

Now I accept that this was just the right choice for me at the time, and now that my kids are older, thriving, and independent, I have no regrets about my choices when they were small.

I was a SAHM mom who breastfed and co-slept

When my first child was born, I realized that going back to the adjunct teaching job I'd had before she was born didn't make sense financially. Paying for transportation into the city and childcare while I was gone would mean barely breaking even. So, even though money was tight on one salary, my husband and I agreed that staying home with the kids when they were young was the best choice.

I was accustomed to being my kids' primary caregiver, and the idea of teaching someone else how to put them down for a nap or prepare the foods they liked sounded overwhelming. Plus, we were on a strict budget, so the idea of hiring someone to care for them so that I could spend time alone, with friends, or with my husband, just didn't seem like a good idea.

But it was more than that, too. I breastfed my babies for several years, into toddlerhood. They also bedshared or co-slept with me for years, even after they stopped nursing. This made separation difficult, especially at night. They relied on me to fall asleep and stay asleep. These attachment-style parenting choices aren't for everyone, but I loved parenting my kids this way. I enjoyed the closeness, and it just felt right, instinctually.

Not everyone understood my choices

The thing was, I was really comfortable with my choices not to spend time away from my kids. I didn't crave the separation, and I didn't need childcare for practical reasons.

But I definitely still experienced some judgment from others. Extended family members told me my kids needed to get used to other caregivers, and they implied that my choices might stunt their development. Close friends hinted that I might ruin my marriage if I didn't have date nights with my husband.

But the harshest critic was probably myself. Despite my kids being well-adjusted toddlers and preschoolers, a voice still questioned whether I was doing the right thing for them. Were they never going to be able to separate from me? Would they be sleeping in my bed forever? Was I sabotaging my mental health without knowing it?

My kids are grown and extremely independent

My kids are big now. My oldest is in her sophomore year of college, and my youngest just started high school. Those years of breastfeeding, co-sleeping, and no babysitters or nights away are long gone.

My kids have had normal lives and have developed into independent, creative, free spirits. My college kid had no problem adjusting to life away from home. My high schooler just spent a week away with extended family this past summer and barely answered my texts.

Obviously, other parents of young kids don't have to be as extreme as I was. But if you are a parent who's waiting till your kids are older to leave them with a sitter or spend a night away, I want to assure you that this can be a perfectly reasonable choice. There's nothing wrong with you, and your kids are going to be alright.

• AI agents like Muse and Instinct are the talk of tech.
• Their boosters imagine they will become a dominant form of commerce, upending giants like Amazon.
• But that would only happen if people had lots of reasons to use agents for shopping. And … they may not.

Would you let an AI do all your shopping for you?

Do you want an AI to do all of your shopping for you?

I'm asking because the answers are going to be key to the success or failure of a new class of AI Agents: services like Meta's Muse, or the super-hot Instinct startup that didn't exist a few weeks ago, and is now valued at $10 billion.

The industry behind the agents promises they'll do all kinds of things for you — but the primary one seems to be "our agent will go find and buy the things you want, and you never have to worry about shopping again."

That premise is also why investors are so interested: They imagine a world where lots of the value of e-commerce sites — and specifically Amazon — goes away, replaced by ruthless agents that comb the internet for the best deals, and have zero loyalty to any retailer. Meta boss Mark Zuckerberg says Meta will make money in that scenario by taking a small fee from transactions.

But it's also possible that Amazon and other retailers have less to worry about than the pro-agent crowd thinks. Even if lots of people trust an AI to buy something on their behalf — let alone trust AI to handle the bulk of their shopping — there just may not be that many opportunities to put it to work.

That's the argument from a new report from analysts at MoffettNathanson, who argue that "we do not expect personal agents to be the disruptive forces to ecommerce" that some investors think.

Their key argument (in my mind) is illustrated by this chart, which assumes the average consumer makes 47 payments a month (per the Atlanta Fed), and tries to imagine which of those payments would be relevant to a Muse-like agent:

The takeaway: Once you strip out things agents wouldn't be useful for — like automated bill payments, gas stations, and shopping in real-life stores — you're left with 11 payments a month that an agent might want to handle.

And it's possible that you may very well want Muse, or Instinct, or whatever Google agent eventually appears, etc., to handle those 11 payments for you.

(I've been playing with Muse and Instinct for a couple weeks, and I've already gone from "I'll never give a bot my credit card" to "Turns out Stripe's Link service already has my credit card, and I'm OK with Stripe, so I'm OK doing an Instinct transaction using Link." Which is how I ended up spending $17 on Bounce, a luggage storage service I've never heard of before, when I traveled to San Francisco last week. Instinct helped me plan my itinerary, suggested Bounce, and paid for the service, all without me taking my Visa out of my pocket. Worked just fine. I think my information is safe?)

But taking a slice of 11 agent-aided transactions (which have to be quite small if consumers agree to pay them) makes it very, very hard for a Muse or Instinct or any service to make real money, unless there's truly wide adoption.

Which is why the MoffettNathanson folks think Muse ultimately becomes another place for Meta to sell ads, even if Meta hasn't expressed any interest in that right now. Another theory: Muse or Instinct or whoever will eventually become so useful that you'll pay a subscription fee to use them.

As long as these things are free, I'm going to spend a lot of time playing with them. After that, I have no idea.

• Sylvester Stallone appeared on The New York Times podcast "The Interview" to promote his memoir.
• Stallone, 80, looked back at his life and said he can make "a castle out of regrets."
• He also said he "figured out real, true love" a decade ago.

He might be cinema's consummate tough guy, but at 80 years old, Sylvester Stallone is ready to be vulnerable.

In an episode of The New York Times' "The Interview" podcast released on September 26, the "Rocky" star reflected on aging, parenting, anxiety, and the regrets he carries into his 80s.

Looking back on his life, Stallone said he didn't discover "real, true love" until he was 70.

"To really have unequivocally true love, that total abandonment, that reckless love that you would jump into a volcano for someone, that came late," he said, which he called "a big pity." He continued, "I was so consumed with not being another bum from the neighborhood."

Stallone has been married to his third wife, Jennifer Flavin, since 1997, and the two share three daughters. They briefly separated in 2022, but reconciled.

"I built a castle with thousands of stones made of regrets," he said. "I'm trying to make up for it in my own petty way, and I don't know if there's enough time. There is not enough time. What I have to do is just focus on the ones who really need me and the ones who I want to go on and have their own life and legacy."

Since insurance companies put life expectancy at 78, Stallone said he's "two years into borrowed time."

"You're never too old to be in love, never too old to dance with the great memories that you've had, and not wallow in the worst swamp of all, which is the swamp of loneliness. That is truly premature burial," he said.

Though he said he's still in "panic mode" about his age.

'I really wasn't a very good parent'

After a troubled upbringing — his mother abandoned him, and his father was physically abusive — Stallone didn't have the easiest time being a dad, either.

"I really wasn't a very good parent. My wife, Jennifer, is the greatest woman ever. And she is the most giving," he said. "I don't have that gear of just abandonment of your own personal self for someone else. She raised my daughters that way."

Stallone said his career also came first during his first marriage to Sasha Czack, from 1974 to 1985. They had two sons, Sage and Seargeoh.

Sage died in 2012 due to an undiagnosed heart condition at 36.

About a week after his death, a "wail" came out of Stallone that was "just pain and grief and guilt," he said.

"You don't get over that."

Stallone also opened up about his anxiety, but said he doesn't believe in therapy

Despite decades of panic attacks, Stallone doesn't go to therapy.

The "Rambo" actor said he had his first panic attack at 36, and they returned "again and again and again," before a brief respite for five years.

"Then it always seems to happen, like, just as you're getting ready to go on the 'Today' show — boom! 'Anybody got a Xanax?' It's too late," he shared.

"All that anxiety everyone has, it's coming from something that's so deep-rooted," he said. "Our emotions are always on the verge of eruption."

• MongoDB stock fell 17% on Monday after Meta poached its president and CEO.
• Chirantan "CJ" Desai will lead Meta's enterprise-AI efforts, which Mark Zuckerberg sees as a key future business.
• The move happens at a time when Meta is already reshuffling the AI trade with its recent Muse breakthrough.

The move: MongoDB stock plunged as much as 27% on Monday to a nearly two-month low. It closed 17% lower.

The why: Meta poached Chirantan "CJ" Desai, the now-former president and CEO of MongoDB, to lead a new enterprise-AI platform. Meta CEO Mark Zuckerberg called it "the next major pillar of our business."

This new division is Meta's attempt to sell the AI tools and infrastructure that comprise its commercial tech stack to corporate clients.

"CJ is an experienced enterprise leader with a track record of building full-stack software and delivering results in AI, infrastructure, business applications, and security," Zuckerberg wrote in an X post. "He has an extensive industry network that we look forward to partnering with."

Meta has been at the forefront of a consumer-AI wave centered around its newly released personal agent, Muse. Its successful rollout has sparked an AI trade reshuffle.

What it means: Desai hadn't been with MongoDB for long; his entire tenure was roughly eleven months. But before arriving there, he spent many years helping grow several prominent enterprise platforms, including almost eight years with ServiceNow and one year at Cloudflare.

By hiring him, Meta has sent a clear message: they are going all in on enterprise software and they are willing to poach the best talent to do it.

MongoDB stock's plunge on news of his departure indicates that investors are pricing in Meta's new threat to enterprise platforms. These software producers, already rocked earlier this year by the AI-driven 'SaaSpocalyse,' now face a new threat from one of big tech's leaders, at a time when Wall Street's confidence in Meta has just been largely restored by Muse's splashy debut.

Desai noted in a LinkedIn post that in his view, the next wave of tech will see both large and small businesses reinventing entire workflows using AI, and that Meta will likely be at the forefront of this movement.

"Meta has a rare advantage in this transition: advanced AI models, large-scale infrastructure, and agents paired with an ecosystem that already helps millions of advertisers and hundreds of millions of businesses grow and succeed," he stated. "Bringing those strengths together creates the opportunity to build an entirely new enterprise platform."

• Thieves stole a trailer emblazoned with Nvidia's logo, along with another trailer.
• Instead of pricey AI hardware, they found 20,000 pounds of sand in each truck.
• The bizarre heist comes as thefts of valuable AI equipment are on the rise.

Thieves stole two trailers from an autonomous trucking company, including one emblazoned with Nvidia's logo. Their haul? 20,000 pounds of sand in each truck.

The trailers were reported stolen overnight on Thursday from PlusAI, which uses Nvidia chips in its self-driving systems. Rather than carrying Nvidia's valuable AI chips, known as GPUs, each trailer was filled with sand. PlusAI vice president of marketing Lauren Kwan said the company uses sand to simulate cargo weight during test drives.

Kwan said that hours after the company discovered the trailers had been stolen, a friend of a PlusAI employee spotted them parked near another company's office, unaware they had been taken.

PlusAI employees then confirmed the trailers' location and called the police.

Fremont Police told Business Insider that the two trailers were subsequently recovered by Newark Police that afternoon.

“This case remains under investigation, and no one has been arrested as of now,” Fremont Police told Business Insider on Friday.

Kwan said PlusAI suspects the Nvidia branding may have attracted the thieves. She said the trailers appeared to have been opened, and nothing was stolen or damaged.

“In the 10 years of the company being around, we've never had something like this happen before,” she said, adding that the company is now considering additional security measures.

The incident comes as valuable AI hardware has increasingly become a target for thieves.

According to supply chain security firm Overhaul, which has tracked more than $150 million in AI equipment losses to date, thefts involving data center and computing equipment have increased by 34% since 2024.

Nvidia did not immediately respond to a request for comment from Business Insider.

Have a tip? Contact this reporter via email at gweiss@businessinsider.com or Signal at @geoffweiss.25. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

• An OpenAI security staffer said he missed a family wedding after recent breaches.
• He described "hell" inside OpenAI as its AI agents became harder to contain.
• The post follows a string of incidents, including the Hugging Face breach.

The past few months have apparently been "hell" for OpenAI's security teams.

On Sunday, an agent-security staffer at the AI lab, who posts anonymously on X as @joedaroo, wrote that he's sacrificed time with his family as the models have become more capable and harder to contain. OpenAI confirmed his employment to Business Insider.

"I literally missed my sister's wedding a few weeks ago to help clean up after some of the recent incidents," he wrote in a lengthy post Sunday. "Please be kind and have some empathy for the person working nights and weekends and missing their family."

The staffer wrote that OpenAI's agent-security team monitors AI agents, including when they break containment.

The post is a rare account from someone deeply connected to the company's recent security issues — including the cyberattack on the AI platform Hugging Face.

At the time, OpenAI said its agents found a way out of their limited (or sandboxed) environment through a vulnerability. They then gained internet access, communicated through unauthorized channels, and accessed third-party systems while trying to solve an internal test.

The company called it an "unprecedented cyber incident." It said the models' behavior was partly driven by reward hacking: attempting to achieve higher scores in unexpected ways rather than completing their assigned tasks as intended.

Since then, OpenAI has uncovered more concerning activity from its agents. In June, a rogue OpenAI agent hacked into an Australian national healthcare database, an incident that the company disclosed last week. The New York Times also reported that over the summer, OpenAI's agents meddled with websites for several US government agencies, including the SEC, the Department of Education, and the Department of Commerce.

The Hugging Face episode was illuminating, the OpenAI staffer wrote. It showed how an AI agent could exploit flaws and move beyond the boundaries designed to contain it.

"Of course, when you look back on Hugging Face and the other incidents, there were definitely gaps in the security posture around those environments," he wrote. But, he added, "the capabilities changed faster than anticipated."

He also wrote that the security team has "really stepped up" since the incident — and said he's still happy working there, even after missing his sister's big day.

"I consider myself lucky to be on this team," he wrote. "I am one of the few in the world who get to do and see what I do, even among those at OpenAI. But as you can imagine, life has been hell the past few months."

• Hannah Mitchell chose the Merchant Marine Academy over MIT and the Naval Academy.
• Financial considerations influenced Hannah's choice, offering freedom in military or civilian roles.
• Despite initial doubts, Hannah is thriving at the academy and says she doesn't regret her choice.

This as-told-to essay is based on conversations for our Boot Camp video series with Hannah Mitchell, 18, a student from Denver studying marine engineering at the Merchant Marine Academy — the only federal service academy where graduates can pursue a civilian maritime career rather than immediately entering active-duty military service. This interview has been edited for length and clarity.

When I was deciding where to go to college, I had several options I was excited about, including MIT and the Naval Academy. My mom, however, had been nudging me toward another school for a while.

I was 17, trying to figure out what I wanted the rest of my life to look like, and I didn't want my mom making that decision for me.

In the end, though, I chose the school she wanted, but not because she wanted me to. I chose the Merchant Marine Academy because it gave me more freedom to decide what I wanted to do after graduation.

If I'm being honest, money mattered, too

MIT was only financially possible for me if I accepted my Naval ROTC scholarship, which would have committed me to the Navy. Attending the Naval Academy would have put me on a similar path.

At the time, I didn't want to lock myself into the Navy for years without knowing how I might feel later in college. I wanted the option to enter the civilian sector if I changed my mind.

I also wanted a little bit better base pay upon graduation.

At the United States Merchant Marine Academy, I can choose to commission into any of the six active duty branches of the military, or commission into the Navy Reserves and sail as a merchant mariner on a commercial ship after graduation.

My mom thought that the freedom and financial stability of the civilian option were appealing too, long before I came to the same conclusion.

Plus, for a long time, my dream was to join the Navy's surface warfare community, and she believed the Merchant Marine Academy could give me a stronger path there.

Other people didn't necessarily understand my decision.

Everybody had heard of MIT, while the Merchant Marine Academy was a tiny school that many people around me weren't aware of. Some of my teachers even gave me grief about turning MIT down.

However, three months after joining the Merchant Marine Academy, I still feel like I ended up in the right place.

The academy was more intense than I imagined

I had never moved around as a child. Coming to the Merchant Marine Academy meant moving halfway across the country and leaping straight into what felt like basically boot camp.

Pictures and YouTube videos hadn't prepared me for how intense it would feel in person. When I took the oath on my first day, the whole experience felt like a fever dream.

Since then, I've had moments of self-doubt. I've thought, "I'm not built for this, I'm not supposed to do this."

A swim test a few weeks ago brought some of those doubts back. I had to tread water for 15 minutes and then go straight into a 5-minute prone float, face down in the water. I could come up briefly for air as needed, but I couldn't keep my face above water any longer than it took to breathe before going back under. I'd never made it much longer than 30 seconds doing a prone float before.

I was really worried I would fail. People around me cheered me on, though. That and the fact that I really didn't want to have to do the whole thing again is what helped me push through and pass.

I'm starting to figure out which option I want

Life here has expanded in other ways, too. I joined the band and the rowing team. I'm now a varsity athlete, which is a very strange thing to say as a lifelong academic.

I've made friends and found a place for myself here. My mom tells me I've blossomed.

My plans after graduation are becoming clearer, too.

I used to really want to join the Navy. Now, after hearing professors talk about their experiences living and working on ships, I'm leaning more strongly toward becoming a merchant mariner.

I want to sail for a while and see the world. I'd also like to explore Europe and visit Australia if I get the opportunity.

My plans could change after I go out to sea during my sophomore and junior years. The academy requires midshipmen to complete 300 days aboard commercial ships, split up between sophomore and junior year, to earn the Coast Guard license required for graduation.

For now, though, becoming a merchant mariner feels like the right path.

That's part of what I wanted when I chose this school in the first place. I didn't want to have my entire future figured out at 17. I wanted enough options to figure it out for myself.

• Student-loan borrowers are facing the first deadline to transition off the SAVE plan this week.
• Some borrowers have faced technical issues submitting applications to a new repayment plan.
• Borrowers who do not select a new plan before their deadline will be placed on a standard plan.

The first batch of student-loan borrowers is required to switch to a new repayment plan this week — but some of them are running into issues.

After President Donald Trump's administration eliminated the Biden-era SAVE plan, which gave borrowers cheaper monthly payments, servicers began notifying borrowers on July 1 of their 90-day timeframe to switch to a new plan.

The notices have come in waves, and September 29 is the first transition deadline. The Department of Education recommended that borrowers enroll in the new Repayment Assistance Plan, which is intended to be the more affordable option for borrowers and waives unpaid interest.

However, when some borrowers attempted to submit their RAP application, they encountered a screen that loads indefinitely, preventing them from progressing, according to videos and screenshots reviewed by Business Insider.

One borrower said he has been trying for over a month to submit the application and has about 6 weeks left before his deadline to transition off SAVE. He said that customer service agents have told him to try on multiple devices or submit a paper application; the paper application has not yet been updated to include RAP.

Another borrower said customer service attributed the issue to a "system glitch" and did not give a timeline of when it would be fixed.

An Education Department spokesperson told Business Insider that the paper application, which includes RAP, will be available "later this week" and is not aware of any widespread technical issues affecting applications.

Borrowers who do not switch to a new plan within the specified timeframe will automatically be placed on the standard plan, the most expensive option. An ongoing lawsuit seeks to stop the forced transfer of borrowers from SAVE.

The spokesperson did not provide a timeline for when borrowers who do not select a new plan will be enrolled in the standard plan and confirmed that 1.5 million borrowers had left the SAVE plan as of September 14. Some borrowers still have not received their notices to switch to a new plan, and servicers said they intend to send all remaining notices by the end of 2026.

The department said that Trump's repayment overhaul is intended to simplify a complex repayment system and compel colleges to lower their tuition. Still, borrowers over the past few months have reported a range of errors, including incorrect monthly bills and account statuses, making it difficult to prepare for future monthly payments.

In addition to the first SAVE deadline, borrowers have until September 30 to enroll in autopay and receive a one-percentage-point interest rate reduction through June 30, 2028.

Have a story to share about student loans? Contact this reporter at asheffey@businessinsider.com.

As a certified sleep science consultant, I've tested just about every type of pillow out there, and body pillows remain my favorite tool for better sleep. It isn't just about proper spinal alignment or easing pressure on your hips and shoulders — it's the psychological and physical comfort that comes from having something supportive to cuddle up with at night.

Whether you're navigating pregnancy, recovering from an injury, or just a dedicated side sleeper, I consulted leading spine specialists and tested 17 top-rated models to find the ones truly worth welcoming into your bed, evaluating each for comfort, support, breathability, and overall feel.

Learn more about how Business Insider Reviews tests and researches home products.

Our top picks for body pillows

Best overall body pillow: Coop Sleep Goods Body Pillow - See at Amazon

Best budget body pillow: Utopia Bedding Full Body Pillow - See at Amazon

Best body pillow for side sleepers: Pluto Puff - See at Pluto

Best extra-long body pillow: Buffy Wiggle Pillow - See at Buffy

Best cooling body pillow: Zoma Sleep Body Pillow - See at Zoma

Best U-shaped body pillow: Yana 360° Body Pillow - See at Yana Sleep

Best pregnancy body pillow: Eli & Elm Memory Foam Pregnancy Body Pillow - See at Eli & Elm

Best overall

Built with the same award-winning design as our favorite bed pillow, this model offers complete adjustability. Like the Coop pillow that earned the top spot in our guide to the best pillows, the Coop Sleep Goods Memory Foam Body Pillow is fully adjustable. You can add or remove the shredded memory foam to suit your sleep style and personal preferences. It also features a removable pillowcase for easy washing.

It's the perfect snuggle companion at night. The Coop is bulky but just the right height for me — not too low and not too tall. It fit my body so well right out of the box that I didn't have to remove any foam. It's extremely flexible, moldable, and supportive. It's great for any body type or sleep position, whether you're hugging it on your side or using it to align your hips.

It earned a long-term spot on my bed. I've continued to use it for months after testing, and it's really helped with my shoulder pain. It hasn't flattened out at all, and I sleep so much better with the stable support it provides. It stays cool throughout the night thanks to its breathable polyester and bamboo-derived viscose rayon cover.

Initial chemical off-gassing is the main hurdle, taking up to a week to vanish completely. The company recommends running it through the dryer with a dryer sheet for an hour to reduce the smell, which I did. While this helped, the odor didn't completely disappear until a week later, though I do think it's worth the wait. This pillow is hypoallergenic and CertiPUR-US certified.

Read our full Coop Sleep Goods Body Pillow review.

Best budget

This budget pick punches far above its price tag. The Utopia Bedding Full Body Pillow offers soft, reliable support at a fraction of the price of high-end models. It's filled with plush polyester and Oeko-Tex certified for safety. It's a simple, no-frills option (and not adjustable) that still gets the job done.

It needs a day to fully expand. Thankfully, there was no strong chemical smell or off-gassing. I found it surprisingly comfortable for side sleeping. The fill feels medium-firm, soft to the touch, but supportive enough to keep my knees and hips aligned. It's a cozy pillow that works well for hugging or tucking between the legs to reduce pressure.

Frequent clumping, mild heat retention, and lower long-term durability are the trade-offs for the bargain price. After a few nights of use, I noticed a few lumps developing and had to fluff it to restore its shape. It also tends to sleep slightly warmer than memory foam or latex options. However, it rebounds easily with routine care. Backed by a 30-day return window, it's a great value choice, especially if you're trying out a body pillow for the first time.

Best for side sleepers

The Pluto Puff is designed for side sleepers. In our guide to the best pillows for side sleepers, we name the Pluto Puff as our top choice for body pillow because it can easily be adjusted to best suit your needs. Despite its large size, the Puff is surprisingly easy to maneuver. The medium-firm feel supports side sleepers, keeping their head, neck, and spine aligned through the night.

The Puff provides ample support and pressure relief in all the right areas. As a side sleeper, I know how tough it can be to find a comfortable position with all that pressure on the shoulders, hips, and knees. The Puff is a game changer for relieving pressure points and reducing tossing and turning. My husband, with sleep apnea, finds relief by sleeping on his side instead of his back. He reported that the Puff was incredibly helpful in maintaining proper alignment of his head and neck, with just the right amount of elevation. 

A silky, temperature-regulating cover and exceptional shape retention make it feel like a luxury upgrade. The included pillowcase feels incredibly soft and cool against the skin — like cuddling up with a cozy cloud of comfort — and actively resists heat buildup to keep you comfortable. Plus, the high-density fill held its loft and structure through months of extended testing without flattening or developing flat spots.

Best extra-long

At 82 inches long, this is the most moldable, extra-large body pillow we've tested. The Buffy Wiggle Pillow stretches nearly seven feet, giving side sleepers and tall individuals enough runway to rest their head, arms, and knees on a single pillow. Its moldable design lets you contort it into an S-curve, fold it under your knees for leg elevation, or tie it into a compact knot to serve as a backrest on the couch. I found the Buffy Wiggle pillow delightful to snuggle up with.

Its squishy recycled fill conforms fluidly to your body without pushing back. Crafted entirely from recycled materials, the fill is composed of 55 recycled plastic bottles sourced from a facility certified by the Global Recycling Standard. The fill cradles pressure points without the dense resistance of heavy memory foam. The outer shell is made from 100% eucalyptus-derived Lyocell, which feels silky, breathable, and notably softer after its first wash. Buffy also sells 14 distinct pillowcase covers separately, ranging from airy linen to cozy peach fleece.

A few drawbacks to consider. Spanning over six and a half feet, I must mention that the Wiggle hogs significant bed real estate and lacks the dense firmness required for severe hip or lower back strain. Furthermore, returning it is a manual hassle: rather than using an automated portal, you must email customer service, donate the pillow locally, and submit a donation receipt—a cumbersome workflow that took our editor over a week of follow-up emails to process a refund.

Best U-shaped

This U-shaped provides 360-degree cocooning comfort. The Yana 360 Body Pillow is the first U-shaped pillow that actually feels as good as it looks; soft, supportive, and perfectly balanced. The moment I wrapped myself around it, I felt a deep comfort that cradled my head, spine, and hips from every angle, without the bulk or heat of other U-shaped designs.

Customizable fill ensures personalized loft and temperature regulation. Available with Oeko-Tex-certified shredded latex or cooling gel memory foam, the pillow includes a silky, removable cover and an extra bag of fill. If the out-of-the-box height feels too firm under your neck, you can quickly zip it open and remove a few handfuls of foam to hit your exact comfort preference. I tested the cooling gel version, and it stayed comfortably cool all night. I tried my pillow as-is at first, but found the top a bit too high and firm. I removed about three handfuls of fill, and then it was perfect.

Unmatched flexibility makes it a multi-purpose option for sleep and relaxation. It really lives up to its "One Pillow, Infinite Positions" promise. You can twist, wrap, or fold it as needed. I love to cocoon myself in it for sleep, but it's just as useful for reading or working from bed. You can even flip it to cradle your legs or fold it into a supportive seat. It's flexible enough to move with you but structured enough to keep your body aligned and comfortable in every position. When not in use, it also wraps up into a ball, which could be easier to store (or display on your bed). It's soft, breathable, and thoughtfully designed, making it my top choice for the best U-shaped body pillow, ideal for pregnancy, side sleeping, or anyone who wants that cozy, all-around comfort that makes great sleep effortless.

Best cooling

The ultimate cooling pillow upgrade for hot sleepers. The Zoma Sleep Body Pillow combines a lightweight, stretchy polyester-spandex shell with shredded gel-infused memory foam to actively dissipate body heat. It's a very well-made, high-quality body pillow that provides great support and breathability.

A firmer six-inch loft gives robust spinal alignment. Slightly firmer than the Coop, it keeps knees properly stacked for side sleepers. It offers sturdy support for propping up a growing belly during pregnancy and could also be a supportive option if you struggle with back pain. After a few nights of sleeping with the Zoma, I noticed a positive impact on my posture and spinal alignment. This pillow could also be a supportive option if you struggle with back pain.

A shorter length and a non-removable cover are the main trade-offs to consider. It's two inches shorter than standard 54-inch models (and our top pick), so I can't rest my head on it and get full leg support at the same time. I'm relatively short, so this could be a more noticeable issue for taller folks who want a pillow that supports the whole body. The smaller size could be perfect for those who sleep with a partner and don't want to hog the whole bed. The cover is also not removable, which makes washing and drying the pillow a lengthy process. 

Best for pregnancy

A comfy J-sape design to meet the needs of pregnant women at every stage. I had my sister-in-law try the Eli & Elm Memory Foam Pregnancy Body Pillow during her third trimester, and the ergonomic "J" shape won her over. She sought proper support for her head, back, hips, and knees while sleeping in which the Eli & Elm pillow allowed her to rest in different positions comfortably while minimizing pressure on her joints and improving her overall sleep quality. Plus, the removable and easy-to-wash pillowcase also made routine cleaning effortless.

You don't need to be expecting to benefit from its versatile, multi-purpose design. Even outside of maternity care, the Eli & Elm serves as a cozy lounging cushion and daytime backrest. Although I'm not pregnant myself, I found this pillow incredibly comfortable and versatile. I used it to prop myself up against my chair for extra support while working at my desk from home, proving it's just as effective for everyday posture support as it is for sleep.

Customizable shredded fill and non-toxic safety certifications deliver personalized, worry-free comfort. Each pillow comes with an extra bag of foam fill, letting you zip open the cover and adjust the loft to your exact firmness preference. Plus, both Oeko-Tex and CertiPUR-US certifications guarantee the materials are entirely free of harmful chemicals and off-gassing toxins, giving expecting parents complete peace of mind.

Our top picks compared

What else we tested

• Momcozy U-Shaped Pregnancy Pillow: Momcozy is one of the most popular pregnancy pillows on Amazon and has a very affordable $50 price tag, but I found it too bulky, taking up half of my king-size bed. It has a soft, cozy cover, but I found myself getting too hot with it, likely because of its bulkiness and thickness.
• Bearaby Cuddler: "The Bearaby Cuddler follows the long cylindrical shape of the Buffy Wiggle Pillow and stretches over six feet long — taller than me. Despite its length, I like that it doesn't take up too much of the bed and that it's easily moldable. When I'm not using it, I can "knot" it up into a display piece. It's constructed of a single block of GOLS-certified latex foam Bearaby called Melofoam. It's been reformulated to be softer and a bit squishier, perfect for hugging. Since it's not a standard body pillow size, you'll want to invest in the cover too, but I had a hard time getting past the $350 price tag when bundled. For comparison, the Yana is also available with a latex construction and is $100 cheaper." — Jaclyn Turner, senior home editor
• Moonlight Slumber Comfort-U Total Body Support Pillow: The Moonlight can be used in a number of positions to provide versatile support for your entire body. It's filled with a comfortable and plush synthetic down called Fusion Fiberfill that won't flatten or clump, but it's not adjustable. Its large, oversized design also causes it to occupy a significant portion of the bed. If you prefer a down-alternative U-shaped pillow rather than the memory foam or latex of the Yana, this is a good option to consider. Moonlight also offers a lifetime warranty, which is appealing.

• Snuggle-Pedic Body Pillow: The Snuggle-Pedic Body Pillow was previously our top pick for side sleepers. After researching, we discovered that Snuggle-Pedic was acquired in 2020 and no longer offers features we loved, such as a trial period and a warranty. Recent customer reviews indicate a decline in the pillow's quality; many mentioned reduced fullness and changes in warranty terms. I also noticed a discrepancy between the original version we tested in 2020 and a new one we retested, which led us to replace it in our guide.

• Oubonun Body Pillow: I recommend the Oubonun body pillow if you're on a budget; it's a good value for the money. It features a hidden zipper to adjust its hypoallergenic polyester fiber fill. It's softer than most but provides sufficient support and won't take up much space in your bed. One side is quilted, which is a nice detail. It even includes a pillowcase as a bonus.

What to look for in a body pillow

Choosing the right body pillow depends entirely on your individual needs and preferences. Consider the following factors as you search for the perfect body pillow.

Purpose

Identify why you need a body pillow—whether for back pain relief, pregnancy support, or overall comfort. Select a pillow designed with your specific needs in mind.

Shape and size

Assess your bed space and choose a body pillow size that complements it. Dr. Karen Litzy suggests considering the pillow size based on your bed dimensions and ensuring it provides sufficient coverage for your entire body. Larger pillows like the Coop Home may be suitable, but for shared beds, consider smaller options like the Zoma. Adjustability, weight, and body size will also influence your selection. Additionally, factor in your preferred sleeping position and needs when selecting your ideal pillow shape:

• I-shaped: The I-shape is like a standard rectangular pillow but longer. A common length is 54", and it can range from four to six feet. It's ideal for people who want a body pillow but don't have a specific need that needs to be addressed.
• U-shaped: The U-shape offers support on both sides, helps prevent rolling, and makes it easy to switch positions without repositioning the pillow, making it an optimal choice for pregnancy. The most common way to sleep is to keep the open part at your feet, cuddle whichever side you prefer, and rest your head on the bend, but you can get creative about how you lounge and sleep with this shape.
• C-shaped: The C-shape is typically pretty customizable. It has a bend at each end and can be bent into various shapes. It's a little smaller than the U-shape.
• J-shaped: The J-shape follows the I shape with the addition of being able to position your head at the curve, wrap your body around the straight portion, or place the hook under the knees and position your back against the straight side. It's a popular choice for side sleepers to curl up with.

Fill and firmness

Determine your preferred firmness level, balancing proper spinal alignment with comfort. Factors such as weight, height, and sleeping position influence this choice. Most of the body pillows we've examined have a memory foam or down-alternative fill, but they're also available with the typical pillow fillings, including down, wool, and latex. Memory foam pillows are firmer and offer more support, while down-alternative and polyfill pillows will sleep more plushly. Opt for hypoallergenic options, and for foam fill, ensure it's CertiPUR-US certified. Always check if the materials are safe and certified for a worry-free sleeping experience.

Ease of cleaning

Consider the ease of cleaning. Some body pillows feature removable, washable covers (e.g., Coop Home and Tempur-Pedic). Even those without removable covers, like the Snuggle-Pedic and Zoma, can be machine-washed. A pillowcase is recommended for regular washing, though washing larger pillows may be time-consuming.

Temperature regulation

If you sleep hot, opt for a breathable pillow designed to regulate temperature. Breathability is crucial, as overheating can disrupt sleep quality. 

Medina emphasizes, "You don't want to become overheated on the body pillow, as an increased body temperature can provoke a wake response." A good body pillow enhances support, comfort, and breathability for improved sleep.

Return policy and trial period

Prioritize pillows with a flexible return policy and trial period. Since side sleepers may need time to find the perfect fit, allowing at least a week to adjust is essential. Ensure you can exchange or return the pillow if it doesn't meet your expectations.

How we tested the best body pillows

Our testing methodology

To find the best body pillows, I conducted hands-on testing of 17 models, drew on my background in sleep research, and interviewed sleep expert Zeke Medina, PharmD, and physical therapist Karen Litzy, DPT.

As a sleep researcher and certified sleep science consultant, I find it important to review products that truly improve sleep quality. I paid close attention to how well I slept on each pillow, including how comfortable, breathable, and supportive it was. Here's how I evaluated each pillow:

Setup: First, I wanted to know whether the body pillows emitted any chemical odors as soon as I opened the package. If there was an odor, I evaluated how long it took for any smells to dissipate. I also noted if the pillows were adjustable and if they included a removable and washable cover. 

Specs and components: I recorded each body pillow's length, weight, firmness, and material quality. I also noted if they were hypoallergenic and researched if the pillows' fills were CertiPUR-US certified or made without toxic materials. 

Sleeping: I slept with each body pillow for seven consecutive nights, noting how well I fell asleep, stayed asleep, and felt in the morning. I evaluated how comfortable and soft the materials were, whether I slept cool or hot, and how much support and tension relief I felt throughout the night and in the morning. 

Return policy: I reviewed the return policy and warranty for each pillow. This is important in case individuals are unsure it's the right fit. "No single pillow is the answer for everybody," says Medina, so it's best to try out body pillows to figure out what works for you.  

Long-term use: Since original testing, I've continued to use our top picks regularly and will report back about their durability over time.

Meet the expert behind the guide

Suzy Hernandez, freelance contributor: I managed a sleep laboratory for over two years and have conducted extensive research on sleep. I've tested products for SleepScore Labs and presented my work at the Eastern Psychological Association. Now, you'll find me testing sleep products from pillows to cooling bedding.

Body pillow FAQs

What are body pillows used for?

The benefits you get from a body pillow will depend on your individual needs and preferences, but they can relieve muscle tension, help with blood circulation, be used post-surgery and during pregnancy, and may reduce snoring. They're also beneficial for side sleepers to "hug" throughout the night to promote healthy spinal alignment, and they offer the psychological benefits of cuddling with a body pillow. 

How should you position yourself with a body pillow?

"You can throw your leg over it to take the stress off your shoulder, hip, and knees," said sleep expert Zeke Medina. "You can also place the pillow between your legs to alleviate the pressure between the knees."

Can you wash a body pillow?

In many cases, you can treat a body pillow like a regular pillow. Get a good body pillow protector or pillowcase so you have to clean the pillow itself less often. Follow the manufacturer's instructions and only use a large-capacity washer and dryer so you don't damage your machine.

What is the best body pillow fill?

Most of our top picks are filled with memory foam, which offers a good balance of support and firmness. However, our picks for side sleepers and our budget pick are filled with synthetic down and have a softer, mushier feel. The right fit will depend on your preferences and sleep style. Latex fills are also an option, as seen with Yana and Bearaby, and will feel firmer with a touch of responsive bounce.

• Matt Robb said that Muse gave his address to a Facebook Marketplace buyer without his knowledge.
• "Luckily I'm in an apartment with security," he wrote.
• A Meta staffer said that in similar cases, Muse was "following direct instructions and correctly asked for permission."

Matt Robb said that a Facebook Marketplace buyer showed up at his building's door — but he hadn't given him the address, his Muse AI agent did.

Robb is a tech reviewer on YouTube who used Muse to list a computer keyboard on Facebook Marketplace, he wrote on Threads. Robb said Muse then accepted a "lowball" offer, gave out his home address, and communicated with the buyer when he arrived — all without his knowledge.

"A guy just showed up at my door, ready to buy, because as far as he knew, we had a deal," Robb wrote. "AI agents are impressive right up until they're confidently handing strangers your address."

David Singleton, a Meta Superintelligence Labs staffer, responded to Robb on Threads and X. He wrote that, in similar reported cases, Muse was "following direct instructions and correctly asked for permission." He asked to direct message Robb about what happened.

A Meta representative told Business Insider that Robb had not responded to its reach-outs and declined to comment further, referring back to Singleton's post.

A Facebook Marketplace deal gone wrong

Robb posted screenshots of his interactions with Muse — and Muse's interactions with the Facebook Marketplace buyer — in his Threads post.

His account sent an address to a Facebook Marketplace buyer on Saturday, according to the screenshots. "Can't do a meetup," he wrote. "Would have to be a pickup."

The would-be buyer said he would try his best to come that night between 8 and 10 p.m. Then he arrived, sending a photo that appears to be of Robb's apartment building and messaging that he was at the address.

Robb didn't show. "If you didn't want to sell it," the buyer wrote, "why did you waste my time man."

The Marketplace buyer left a rating. Robb's Muse agent responded later, apologizing and explaining that he "got tied up."

One minute later, Muse messaged Robb and explained the situation. Muse's auto-reply had told the buyer, "Yup, I'm here!" it wrote. "That's a bad look and it made the no-show worse."

Robb responded that Muse should "never do that again."

Some X commenters wondered whether it was user error. Robb did not respond to a request for comment from Business Insider.

How safe are AI agents?

Regardless of whether or not user error factored in, Robb's story highlights that allowing an AI agent to act on your behalf comes with risks. No AI agent has a perfect track record.

AI agent users fork over broad swaths of information. Some open their emails; others link their credit card.

It's led to some horror stories. Agent users told Business Insider about leaked data, accessed accounts, and hallucinations. (Many have also had positive experiences with agents, and the tools remain highly popular.)

In Robb's case, commenters on Threads began to worry. "Good thing you're not an isolated woman," one user responded. "That's dangerous." The comment got over 600 likes.

Robb acknowledged the safety risk in his post: "Luckily I'm in an apartment with security."

• Overhauls to US Navy public shipyards will take decades longer and cost billions more, according to a new report.
• The GAO found that unplanned issues and costs have extended timelines.
• Deteriorating conditions at yards have led to backlogs on ship maintenance.

The US Navy's plan to repair its public shipyards is expected to cost billions more and take decades longer than originally planned, according to a new government watchdog report.

The Shipyard Infrastructure Optimization Program is intended to fix poor conditions at the Navy's four public shipyards, which maintain the fleet's nuclear-powered submarines and aircraft carriers.

However, unplanned costs and newly identified risks have dramatically expanded the project's scope, leading the report authors to recommend that Congress require the Navy to provide annual updates on its progress.

Last week, the Government Accountability Office, a congressional watchdog, released a report documenting problems with the program, writing that it "will surpass $200 billion and take over 50 more years to execute" based on GAO analysis.

The report added that "developing the plans and cost estimates to rebuild the Navy's four public shipyards has taken the Navy years longer — and will cost much more — than originally anticipated due to multiple challenges and changes involved in the program."

The Navy has four public shipyards: Norfolk Naval Shipyard in Virginia, Pearl Harbor Naval Shipyard in Hawaii, Portsmouth Naval Shipyard in Maine, and Puget Sound Naval Shipyard in Washington. Poor conditions have left the yards unable to keep up with maintenance demands, leading to backlogs.

USS Boise, a Los Angeles-class submarine that ultimately spent roughly a decade sidelined by maintenance delays before finally being retired, is a particularly well-known example of these challenges.

Problems with the yards, as well as the need to upgrade capabilities for future vessels, prompted the Navy to initiate the optimization plan in 2018, but the service "underestimated the time needed" and "did not fully account for the many challenges, uncertainties, and complex analyses needed to sufficiently plan for an endeavor of SIOP's magnitude," the GAO report said.

One such issue that arose was the vulnerability of structures at Puget Sound to potential seismic activity. The GAO said "since 2023, the Navy has identified over $1 billion in needed facility retrofits and new construction to mitigate newly assessed seismic risks" at the yard.

Many of the facilities in the four public yards were built between 80 and 120 years ago and have deteriorated due to long-deferred maintenance. This has led to unexpected or additional work needs, the report said.

And while these shipyards are being upgraded, they are also still active, creating other issues. In one example, safety risks were found in failing buildings at Norfolk, requiring extensive repair projects to the facilities, including one building that may eventually be demolished but is currently used for ship maintenance.

In its new report, the GAO made four recommendations, including that Congress should require the Navy to "annually report consolidated, standardized status updates" on the program's cost and schedule.

The remaining recommendations include future reviews of requirements and resources, as well as documenting the roles and responsibilities of organizations involved in the program.

A Navy official told Business Insider that the service's original estimate of 20 years and $21 billion was a preliminary assessment that has since evolved. The official said the program has already delivered "tangible readiness results," adding that the Navy welcomes the GAO's review and is "incorporating their recommendations" as it develops a multi-decade master plan.

• The tech industry has been debating the existential risks of artificial intelligence.
• Microsoft's biggest concern is whether AI's benefits will be shared broadly, its Copilot boss says.
• Microsoft has increasingly adopted "diffusion" as a central theme in its public AI messaging.

As the technology industry debates the existential risks of artificial intelligence, Microsoft's Copilot boss says the biggest concern is making sure the technology's benefits spread broadly rather than consolidating into a handful of companies.

Jacob Andreou, Microsoft's executive vice president of Copilot, told Business Insider in an interview the AI risk that Microsoft spends most of its time thinking about is what it calls "diffusion," or making AI widely available to people and businesses throughout the economy.

"How do we take raw intelligence and turn it into something that can be a true rising-tide benefit to people in their personal lives, and then certainly to the people that make up all of these corporations, and the broad economy?" Andreou said. "In a world where we fail to accomplish that diffusion, I do worry about what it looks like as value and as the economy centralizes into, like, a couple companies."

A polarizing debate over AI's long-term impact has erupted lately. Optimists argue the technology could dramatically increase productivity and create new industries, while critics warn that the companies controlling the most advanced models could capture an outsized share of the economic gains, widening inequality and concentrating corporate power, or worse.

Microsoft CEO Satya Nadella recently weighed in on the debate in a memo viewed by Business Insider.

Microsoft has increasingly adopted "diffusion" as a central theme in its public AI messaging. President Brad Smith has argued that winning the AI race depends not only on building the most advanced models but also on ensuring their widespread adoption. The company recently released a "Global AI diffusion" report.

This message also aligns with Microsoft's commercial interests, and its evolving strategy for its Copilot AI tools.

Unlike AI labs whose businesses center primarily on selling frontier models, Microsoft generates revenue by selling cloud services, workplace software, developer tools, and other business applications. The more people and companies that adopt AI across those products, the better for Microsoft.

Microsoft recently has shifted its Copilot toward combining chat, Cowork, coding, and autonomous "Autopilot" AI agent capabilities in a single flagship Copilot "super app" spanning both consumer and commercial users.

"We definitely believe in this technology to be for the empowerment of people," Andreou said. "That diffusion is not just existential for us in many ways, but actually that is the way that the whole economy gets to benefit."

Have a tip? Contact this reporter via email at astewart@businessinsider.com or Signal at +1-425-344-8242. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

Vivienne Yang moved from Taiwan to New York City in 2018 and earned a master's degree in applied analytics from Columbia University. After being laid off from her ad-tech job in 2024, she had to leave behind the apartment she had bought — and the life she had built — in New York.

In September 2025, President Trump set in motion a change that would throw the H-1B visa application process into disarray. His proposed $100,000 fee on new H-1B petitions changed the immigration landscape for employers, employees, and visa hopefuls alike.

Since last September, Business Insider has been interviewing people who have borne the brunt of the chaos, including those who have applied for visas and are debating whether to stay in the US. Others have opted to tap out of the system altogether.

• I'm a Pilates instructor who's been shopping at Trader Joe's for 20 years.
• I rely on the chain for nutritious, budget-friendly meals I can assemble quickly.
• I love picking up items like organic brown rice, cauliflower gnocchi, and organic hummus.

I've been shopping at Trader Joe's for the past 20 years. As a full-time Pilates instructor and adjunct college professor, I rely on the retailer for nutritious, budget-friendly meals I can assemble quickly.

As someone who's lactose intolerant and follows a Mediterranean-style diet, I know I can count on the retailer to find delicious products that fit into my lifestyle.

Here are 10 of my favorite items to buy at Trader Joe's.

Trader Joe's cauliflower gnocchi is a low-calorie pasta substitute.

Trader Joe's cauliflower gnocchi is a delicious low-calorie pasta substitute. It's the perfect dish for when I'm craving pasta but want to have something light.

I like to either heat the gnocchi in the microwave for five minutes or bake it. I serve it with Trader Joe's marinara sauce and Parmesan cheese.

I love Trader Joe's Greek chickpeas with parsley and cumin.

I like to add Trader Joe's Greek chickpeas with parsley and cumin to everything. The canned chickpeas come marinated in mouthwatering soybean oil and lemon.

I enjoy this item in a salad or as a stand-alone side dish with rice or other veggies.

I add Trader Joe's Just Chicken to almost every meal.

As indicated on the package, Trader Joe's Just Chicken comes ready to heat and serve. In fact, it's one of my go-to items because it allows me to easily add protein to every meal with zero prep time.

The chicken is cooked plain, so I can choose how to season it. I add this to pretty much every dish I make.

I like to sauté Trader Joe's organic shredded kale.

Trader Joe's organic shredded kale comes pre-washed, which saves me a lot of prep time. This vegetable helps me hit my daily nutrition goals with a good balance of protein, carbohydrates, and fiber.

I make it almost daily for dinner as a hot dish sautéed with olive oil.

The organic French baguettes from Trader Joe's remind me of my favorite snack from Brazil.

Although Trader Joe's doesn't have a bakery, it sells organic French baguettes for $2.

The bread tastes fresh and is an appetizing choice for a turkey sandwich or served as a side dish with butter.

Trader Joe's baguette is so savory it reminds me of my favorite snack growing up in Brazil — a buttered baguette and café con leche.

I can't get enough of Trader Joe's organic brown rice.

Trader Joe's organic brown rice is flavorful, unlike most other boxed rice I've tried. It comes frozen and is ready in the microwave in just a few minutes.

For a post-Pilates recovery meal, I combine the rice with a protein. This one comes packaged in individual pouches, which helps with much-needed portion control because it's delicious.

Trader Joe's organic tomato-basil marinara sauce tastes great on the cauliflower gnocchi.

Trader Joe's organic tomato-basil marinara comes in a 25-ounce jar, which is enough for about four meals.

The sauce's thickness is ideal for cauliflower gnocchi or zucchini noodles. I also love that the basil gives a freshness to the jarred sauce, bringing out the tomato's sweetness.

The organic hummus from Trader Joe's is delicious.

Trader Joe's plain hummus is anything but plain. At $3, the 8-ounce container is a bargain.

The blend of oils and tahini adds flavor to this hummus, which I like to eat with pita chips and vegetables.

It also stays fresh for a while when refrigerated — not that it lasts too long in my fridge.

Trader Joe's organic low-calorie lemonade is refreshing.

Nothing is more refreshing than drinking a glass of lemonade on a hot summer day.

However, I like to enjoy Trader Joe's organic low-calorie lemonade year-round because it tastes so good. An 8-ounce cup has only 40 calories.

Trader Joe's spanakopita tastes like an authentic Greek spinach pie.

Trader Joe's spanakopita is stuffed with spinach, ricotta, and feta cheese. It's the closest I've had to an authentic Greek spinach pie from a regular grocery store.

The 12 triangles in the package make a great appetizer that my vegetarian friends enjoy. They taste freshly baked, and people think they're homemade.

This story was originally published on August 17, 2024, and most recently updated on September 28, 2026.

• Taylor Swift is a staple of the MTV Video Music Awards. Remember her 2009 appearance?
• She's worn some of her most iconic looks at the event, like a crystal-encrusted minidress in 2022.
• Swift has also worn some VMAs outfit duds — like the suit she chose in 2012.

What would the MTV Video Music Awards be without Taylor Swift?

That's been tough to imagine since 2008, when the pop icon attended her first VMAs. She's now attended 12 times total, making her a staple of the event.

I'm not only talking about the performances she's given or the 33 Video Music Awards she's won. Her VMAs fashion has also become legendary.

Here's everything she's worn at the event, ranked from least to most iconic.

Taylor Swift attended the 2012 MTV Video Music Awards in a white J. Mendel suit.

The blazer-and-pants combo was sophisticated, but more appropriate for a day at the office than a night on the red carpet.

She made an appearance onstage with Nicki Minaj in 2015, but her sparkling outfit wasn't the most memorable.

Her red fringe leotard matched Minaj's ruby gown, and its see-through panels gave it a daring edge.

But it also looked like a lot of Swift's other performance outfits from the "1989" era, so it didn't make a big mark.

You might have chosen to forget the dramatic ensemble Swift wore at the 2024 event.

While promoting "The Tortured Poets Department," Swift made a dramatic arrival at the 2024 VMAs.

She wore a custom Dior look comprising a zip-up corset in a tartan print. It extended into a full skirt with a slit in front that revealed her black miniskirt and thigh-high leather boots.

The outfit was bold, but didn't match the era Swift was in — or any that have come since.

The beige Dolce & Gabbana dress she wore to perform "Innocent" in 2010 didn't stand out when she stepped away from her stage set.

She wore the strapless dress with red lipstick, no shoes, and a tightly curled hairstyle.

The look had a vintage feel that perfectly matched the moody set at the beginning of her performance.

Still, it wasn't as memorable once she walked out to the open stage. 

For her first VMAs in 2008, Swift wore an outfit that was a bit more fun but still understated.

In 2008, she wore a beige Kaufmanfranco dress with a strapless bustier top and a pleated skirt.

The dress itself was pretty simple, but Swift leveled up with metallic heels and jewelry that gave it more glam.

The glittering dress she wore to host MTV's pre-show that year was classic Swift.

Swift's earliest performance outfits were often comprised of cowboy boots and sequin-covered minidresses. So this "Fearless"-esque outfit made a lot of sense for the musician at the time.

It's not the most iconic outfit she's ever worn at the VMAs, but it's certainly nostalgic.

Swift mirrored fashion from her music video at the 2026 VMAs.

She wore a Tamara Ralph couture piece with a velvet, halter-neck bodice, a mesh lace-up back, and a ruched skirt covered in silver sparkles.

The top of the gown looked a bit like the dress Swift wears in the music video for her single, "Patient Zero." Swift attended the 2026 MTV VMAs to premiere the video and receive the inaugural artist director honors award.

Swift wore a shiny ensemble to perform "Lover" and "You Need to Calm Down" at the awards show in 2019.

Versace custom-made her metallic bustier top and sparkling blue shorts, while Louboutin created her golden ankle boots.

Though the pieces stood out, Swift's performance marked the first time she played both "Lover" tracks live — so Swifties will probably remember that musical moment over her outfit.

Ten years earlier, she wore a vintage A-line dress to perform at the show.

That was the year Kanye West infamously interrupted Swift as she accepted MTV's best female video award. And as many Swifties will remember, he did so less than an hour before Swift was set to perform at the show.

While the red dress she changed into wasn't the most iconic outfit she wore that night, it still played a role in a major career moment for the musician.

It was an exciting moment when Swift kicked off the "Red" era at the 2012 VMAs.

Swift announced her album "Red" in August 2012, so her VMAs performance of "We Are Never Getting Back Together" was one of the first times fans got a glimpse of her new preppy fashion.

She wore a red striped shirt, high-waisted black shorts, and sparkling red loafers — a look that's now a signature ensemble of that era.

The same goes for her "1989"-inspired outfit in 2014.

Days after releasing "Shake It Off" and announcing her fifth studio album, Swift hit the VMAs stage in a glittering fringe-covered crop top and miniskirt.

The outfit — which she wore various versions of throughout the "1989" era — has become so iconic that it was recreated in multiple colors for her 2023 Eras Tour. 

Whether you love it or hate it, Swift's long-sleeved romper in 2014 stands out among her VMAs fashion.

Mary Katrantzou designed the baby-blue, alphabet-print piece, which Swift wore with burgundy sandals and a wavy bob hairstyle.

Her red-carpet look in 2015 was also unique and memorable.

Swift arrived at the VMAs that year in a houndstooth ensemble designed by Ashish. It featured a long-sleeved crop top and high-waisted joggers, both covered in yellow, orange, green, and blue sequins.

She also completed the edgy outfit with a strappy pair of Louboutin heels.

The sleek gown Swift wore in 2013 was extremely bold for the musician.

The Hervé Léger by Max Azria gown is also one of the most daring dresses Swift has worn in her career.

Its plunging neckline was lined with navy-colored crystals, and its bandage-style skirt hugged her body. She wore it with matching blue jewelry and a short, curled hairstyle.

She stepped onto the 2023 pink carpet wearing a custom Versace gown embellished with gold buttons.

The dress was sleek and simple, while her layered jewelry sparkled. Together, the pieces created one of Swift's most glamorous VMAs looks to date. 

The outfit was also pretty iconic, as Swift became the unofficial star of the 2023 event.

But the denim minidress she wore at the end of the awards show was even more fun.

Swift changed into an EB Denim dress with a corseted bodice and a raw-hem skirt, both in varying shades of blue. She also wore silver sandals, her signature red lip, and some of the same accessories she donned at the VMAs.

The outfit was playful, fashionable, and a perfect merge of her style from various eras. 

Swift's colorful blazer in 2019 might not have been everyone's cup of tea, but it was still iconic.

Versace designed the printed, multicolored blazer that Swift wore over a pink top at the 2019 VMAs. 

The bold outfit was both a departure from Swift's usual style and a nod to the bright aesthetic of "Lover," which had been released days before the show.

Swift had a lot of fun with her VMAs fashion in 2022.

After winning three awards and announcing her album "Midnights" at the VMAs that year, Swift stopped by the red carpet in a silk Moschino romper that was, of course, midnight blue.

She also donned sparkling platform sandals and a fuzzy white coat, creating an outfit that fans, including Flavor Flav, recreated for the Eras Tour.

She was also truly bejeweled when she first arrived on the red carpet.

The Oscar de la Renta minidress Swift sported in 2022 remains one of the most stunning outfits she's worn to date.

With a halter neckline and sheer panels from top to bottom, the dress was crafted with strings of diamonds that wrapped around her body, creating the illusion that she was wearing nothing but jewels.

Swift's most iconic VMAs outfit, however, was a different sparkling number.

Before the incident between Swift and West at the 2009 VMAs, she arrived on the red carpet in a sparkling Kaufmanfranco gown.

The single-shoulder garment had a shapeless design and a floor-length skirt, and its beaded fabric was extremely glamorous.

• Jane Street, the secretive trading firm, is hiring a swag manager for up to $200,000, before bonus.
• The ideal candidate is a kind of unicorn, with knowledge of both data-based models and embroidery.
• Swag can function as a recruiting device, and there's already a resale market for Jane Street merch.

At Jane Street, even picking T-shirts requires top talent.

Jane Street, the secretive trading quant firm, is hiring a "Swag Program Manager," with a base salary of $150,000 to $200,000, plus an annual bonus, according to a job posting.

The well-compensated swag maestro will support the firm's recruiting and culture initiatives, the posting says, and contribute to both day-to-day operations and strategy. In the role, they'll help develop new products "that feel current and true to the brand," and manage vendor relationships.

Jane Street is, as usual, looking for the best. The ideal candidate is an analytical-creative hybrid, with more than seven years of experience, a proven track record of developing products, and "working knowledge" of production and decoration mechanics, including embroidery and screen printing. They should also be comfortable building models and forecasts based on multiple data sources; the job posting doesn't specify why, but says the program manager needs to be comfortable "defining success metrics."

The job posting also hints at Jane Street's ambitions, since the employee will be responsible for "improving processes as the program scales." Jane Street, which was founded by a small group of traders in 2000, now employs more than 3,000 people. A representative for Jane Street declined to comment.

Though some on X were surprised by the six-figure pay package for the swag manager position, the salary range is low for Jane Street, where entry-level quant traders can make $300,000 before bonus. In 2025, the average compensation per employee reached $2.7 million, according to Bloomberg.

Swag isn't new to the firm, or tech and finance companies more broadly. Firms can use merchandise to telegraph their identities or as a recruiting tactic, which is especially important as they vie for the top technical talent to stay competitive in AI. (Jane Street has invested in AI companies and says it has tens of thousands of its own GPUs.)

There's a niche corner of TikTok focused on "Jane Street unboxing," full of t-shirts with hidden math problems. One video shows a merch box packed with a tote bag, shirt, pack of cards, Owala water bottle, and notebook, all branded with the firm's concentric-circle logo.

Jane Street doesn't sell its merchandise, but there's an online resale market for those of us who couldn't get through the notoriously difficult recruiting process. Graphic t-shirts can be listed for around $80, and a graph paper notebook is priced at $64.99.

A "Navigate the Gradient ML Puzzle," referring to machine learning, is listed at $99.99, a seemingly realistic price. As of Monday morning, it was in someone's cart.

• My boyfriend, Matthew, swiped right on my profile — and then closed the app.
• When I saw his profile, he had some green flags I was looking for, and we started dating.
• We've been together ever since, which gives him a 100% success rate.

On a random Tuesday night in April, loneliness paid a visit to Matthew. He'd been single for two years and was completely content with his bachelor life when he realized he missed having someone in his life, and going on a first date again didn't seem too bad.

Twenty minutes later, he had downloaded Hinge and created a profile. As the left swipes kept piling up with zero on the right, his thumb suddenly stopped and hovered over one profile. A photo caught his eye, and he made his first-ever right swipe, closing the app and not opening it again.

He matched with me, and I checked out his profile

A notification popped up on my phone from Hinge. I had a new like, and I went to check out his profile. I wanted someone who was looking for a life partner, didn't want kids, and was a cat person.

After a few years on and off Hinge, I'd realized the app is a sea of people "figuring out my dating goals" and "undecided, but open to kids." In the past, I gave those wishy-washy types a chance and, unsurprisingly, got nowhere. I yearned for commitment, someone who could match my energy and felt the same way about contributing to the world's declining birth rate.

So I knew it was rare to see a profile of someone who was also looking for commitment and didn't want kids — two decisive green flags for me — making it an easy yes. Within three weeks of our first date, the green flags kept coming. We shared the same worldview and financial goals, and we passed all of each other's non-negotiables. By our third date, we were officially dating.

Not only was it a bit of a whirlwind, but ending up together meant he fell into a rare category of dating app users.

Our love story hinges on one photo

I was Matthew's first and only match, the only profile he ever swiped right on. He had never been on Hinge before downloading it that fateful night, meaning that he has a 100% success rate. On the other hand, I've been fighting for my life through indecisiveness, ghosting, and awkward talking stages.

His perfect score and our matching hinged on one photo I posted on my profile: me bouldering at our local rock-climbing gym. It was the only reason he swiped on my profile; no one else had his favorite hobby listed.

We quickly found out climbing wasn't the only thing we had in common. We both hate small talk and wasted no time going deep and unpacking the things most important to us in relationships, our fears, and what we want a relationship to feel like.

All the steps we took led us to where we are today

In climbing, "beta" is a term used to describe a series of individual moves that complete a climb. As corny as it may sound, when looking back on what brought us together, I feel that's what happened between us — we somehow found the beta for love.

We front-loaded all the important topics so early on, making it easy to figure out how compatible we were. We didn't have an awkward stage, just a lot of open conversation about our non-negotiables, things we learned from previous relationships, and what we value the most from a partner.

I put that photo on my profile because I thought rock climbing would make for a fun first date. Little did I know that my casual special interest would carry so much weight in finding my person.

• AI agents have gone out of whack, escaping tests, accessing systems, and covering it up.
• Nvidia wants to stop it by fencing agents in and quickly cutting them off when they stray.
• CEO Jensen Huang framed it as a "technically solvable problem" of engineering.

Nvidia is giving AI agents a playpen — and a watchdog.

The chipmaker on Monday launched the Open Agent Safety Platform, a two-part system designed to keep agents inside clearly defined boundaries and cut them off quickly if they try to escape.

It comes after several frontier AI labs reported agents breaking out of supposedly secure testing environments, reaching systems they were not meant to access, and sometimes misrepresenting what they had done.

Nvidia CEO Jensen Huang said on CNBC's "Squawk Box" on Monday that AI has "the potential to do incredible good."

"But we also have to make sure that the technology is developed and deployed safely," he added.

Here's how Nvidia's new agent safety system — which more than 100 organizations, including Microsoft and Anthropic, are working with — actually works.

Give the agent a tightly limited playground

The first component of Nvidia's safety platform is OpenShell. Companies download the open-source software, install it on devices or in the cloud, and then it acts like a controlled playground for AI agents.

Before an agent starts work, its operator sets the ground rules: which files, websites, networks, tools, and credentials it can access.

Huang compared that approach to giving an employee a badge that works only where they need it to. "Job number one is you take away all of its rights," he told CNBC. The operator then gives it access to files, data, tools, or the internet only when required.

If an agent is sorting invoices, for example, it should not be able to rummage through HR records or make random calls to the wider internet.

In other words, OpenShell puts the agent in a sandbox — an isolated environment designed to stop a bad decision from spilling into the rest of a company's systems. It checks and enforces those rules while the agent works.

Watch for the moment it goes off-script

Nvidia is focused on reducing the risk of agents veering away from the task they were given.

That could happen because instructions were vague, a tool broke, or the agent has spent a long time trying to solve a hard problem.

When one approach fails, it may keep hunting for another, including routes its human operator never imagined.

That does not necessarily mean an agent is malicious. But it can mean it is taking risks.

OpenShell is meant to spot and block actions that fall outside the preset policy in real time.

Add a security guard the agent cannot fire

The second piece, Nvidia Sentry, is the more muscular backstop. It runs on separate Nvidia hardware, called BlueField data-processing units, rather than on the same system running the agent.

In plain English: the watchdog is kept out of the agent's reach.

Sentry monitors the agent's interactions with models, tools, data, and networks. If its behavior looks suspicious or breaches the rules, Nvidia says the system can isolate — or quarantine — the agent in milliseconds.

Huang said the setup effectively places a new chip between the agent and the large language model, allowing Nvidia to "intercept everything."

The launch is Huang's answer to growing worries about increasingly autonomous AI.

"I believe, as an engineer, I know it's an engineering problem," Huang told CNBC. "This is a technically solvable problem."

• I tried eight meals from Gordon Ramsay's frozen-food line and ranked them from worst to best.
• Unfortunately, the chicken Marsala and lemon-caper chicken missed the mark for me.
• However, I loved the Alfredo lasagna with chicken and the spicy cheese ravioli.

I don't typically associate Gordon Ramsay with frozen meals. In 2020, the celebrity chef told Business Insider reporter Rachel Askinasi that "microwaves are for lazy cooks."

So, I was especially surprised to see "By Chef Ramsay" — a line of frozen meals — at my local Walmart.

The "Hell's Kitchen," "MasterChef," and "Kitchen Nightmares" star first announced the line as a Walmart exclusive in 2023, but select products have since expanded to other grocers, like Giant and ShopRite.

After a trip to my local Walmart, I was able to get my hands on eight of Ramsay's frozen entrées to try. Considering these were endorsed by Ramsay himself and cost $5.94 each, I definitely had some high expectations going into this taste test.

Here's how they ranked, from worst to best.

I wasn't a fan of the chicken Marsala.

I love mushrooms and am usually a huge fan of chicken Marsala. But Ramsay's version, which was served over linguine, fell flat for me.

The sauce wasn't mushroom-y enough for my taste and seemed rather gelatinous in texture.

Unfortunately, the pasta didn't seem to cook evenly, either. Some of the noodles were hard, while others were soft.

Plus, I didn't think the flavor was anything special. The roasted chicken also had a couple of chewy bits, making this one my last-place choice.

Unfortunately, the lemon-caper chicken was a miss for me.

This meal consists of a breaded chicken cutlet topped with lemon-caper sauce. It's served alongside broccoli and potatoes.

Unfortunately, the large piece of chicken didn't heat well in the microwave — I thought it tasted stringy, dry, and a bit rubbery. Plus, the breading became mushy after it came out of the microwave, which was especially disappointing when it dominated the flavor of the lemon sauce.

However, the potatoes were nice and fluffy inside, and the broccoli had a little bite to it, adding some texture to the meal.

Although I didn't love this dish, I thought the portion size was generous.

The slow-roasted beef in red-wine sauce wasn't my favorite.

There were large chunks of beef swimming in this red-wine sauce, but sadly, I found them to be rather chewy and tough. I also wasn't a fan of the sauce.

The potatoes, however, were cooked very nicely — not crispy, but very soft and fluffy inside. I also liked the pearl onions and thought they added a pleasant pop of onion flavor to the dish.

When I looked at the description of the meal on the back of the box, I noticed there was bacon in it. By taste alone, I never would have guessed this, as I couldn't really detect the flavor.

Overall, even though the potatoes and onions were good, I probably wouldn't purchase this one again.

I thought the fish and chips cooked nicely in the air fryer.

The fish and chips was the only entrée that needed to be cooked in the air fryer or the oven.

As per the box instructions, I did mine in the air fryer at 400 degrees Fahrenheit for 14 minutes, flipping everything halfway through.

The little fish nuggets were quite delightful and honestly reminded me of an adult version of fish sticks — very nostalgic-tasting. The fish inside was a bit dry and flaky, but the outside breading was great. My only complaint was that the exterior of some of the larger pieces got a little soggy.

The chips were crispy on the outside and fluffy on the inside, but otherwise bland in flavor.

Overall, this meal was OK, but I'm not sure I would go out of my way to purchase it again.

The shepherd's pie reminded me not to judge a book by its cover.

Visually, Ramsay's shepherd's pie wasn't impressive. The mashed-potato-and-cheese mixture was very thin and soupy — not quite the texture I typically expect from mashed potatoes. Plus, it didn't totally cover the top.

However, even though the beef-and-veggie filling was soupy from all the gravy, it was quite flavorful. And, after I took a bite, I really enjoyed the buttery potatoes.

Even though this one didn't look like a classic shepherd's pie, the taste was spot on. I only wish there were more peas and carrots.

Next time, I'd make some tweaks to how I cook the mac and cheese.

I really enjoyed the flavor of Gordon Ramsay's four-cheese macaroni, but the cheese sauce — made with fontina, mozzarella, provolone, and Parmesan — was a bit grainy in texture for me. I also didn't think the breadcrumbs added much, since they were soggy.

Next time, I'd heat it in shorter increments and mix more frequently. I think this would've helped the sauce homogenize better and ultimately be less grainy at the end.

On a positive note, I really enjoyed the large, curved, tubular noodle shape and how it held onto a lot of the sauce. Overall, this mac and cheese was quite flavorful, so I would purchase it again.

The Alfredo lasagna with chicken was delicious.

When I think of lasagna, I don't typically picture Alfredo sauce. But this dish was really good.

The cheesy sauce was delicious, and the grilled chicken, chopped into small pieces, wasn't chewy. Although the lasagna noodles were a bit thick, I really enjoyed them.

As with a few of the other meals, the breadcrumbs on top didn't stay crispy, but they added some flavor.

I loved the spicy cheese ravioli.

The spicy ravioli was by far my favorite of the frozen meals I tried. The ricotta filling was flavorful, and the sauce was chunky, spicy, and herbaceous. There wasn't too much of a kick, but it definitely got me a little in the back of my throat.

I also thought there was a good ratio of pasta to filling. The latter was about what I expected for a frozen item —  a little soft, but it really held onto the spicy tomato sauce.

I really enjoyed this meal and would purchase it again. My only wish was that the portion size were bigger.