• Kirsten Bradford burned out after spending four to eight hours a day job hunting for over a year.
• She implemented several lifestyle changes after her son brought the burnout to her attention.
• This is the second installment of a 5-part personal essay series, The Burnout Cure.

This as-told-to essay is based on a conversation with Kirsten Bradford, a 30-year-old marketing professional based in Dallas, Texas. It's been edited for length and clarity.

After losing my marketing job in 2023, I had no doubts I'd find a new one. I had just earned my master's degree and figured it would help me break into new industries. To find out how wrong I was, especially as a single parent, was scary and confusing.

Over the next year, I spent every day at home applying for jobs, only to feel exhausted and unable to do much else. I felt heavy and numb, like nothingness. I didn't realize how burned out I'd become until my 8-year-old son brought it to my attention.

I looked around at the laundry piled up, the weight I gained, and my dirty room, and realized all the ways my lifestyle was hurting me. I knew I needed to make a change for us.

Since then, I've completely changed the way I approach the job hunt, and that discipline is helping me recover from burnout and feel like a human again.

In my first year of unemployment, I spent 4 to 8 hours a day applying for jobs

When I got hired for a marketing role at a tech recruitment firm, a lot of the language they were speaking went right over my head. I decided to pursue a master's in business management from SMU Cox to close the learning gap at work. When the school program intensified, and I started pulling myself in too many directions, I was let go from work.

I graduated soon after and started applying to banking jobs, hoping to get my foot in the door. When that wasn't picking up, I figured I'd for sure be able to get a job in marketing given my experience, but nothing came from that either. I just felt so confused. It was scary to go from a full-time salary to nothing, but luckily, my son and I live with my parents rent-free.

In my first year of job searching, I applied to 278 jobs and only landed two interviews. It had me questioning my worth and wondering if something was wrong with me.

It was a really sad time, but I kept pushing through.

My energy was exhausting and my son could feel it

I realized things needed to change when my son, a chronic co-sleeper who is known to tiptoe into my room in the middle of the night, told me, "I don't want to sleep here anymore, I don't like the way it feels."

I knew it was because of me. My energy was exhausting, and he could feel it; no child should bear the weight of their parents' emotions. That's when I told myself, "If you're not going to be happy for you, you need to do it for him."

We always used to have fun together. We'd go to the park, downtown, anywhere to spend time with each other. I couldn't remember the last time we went on an adventure. I couldn't even remember the last time I worked out, cooked a proper meal, or even put on makeup.

I don't think my son recognized me, and at that moment, I didn't either, but I wasn't willing to lose my son over it. My favorite job is being a mom.

I implemented 3 immediate changes in my lifestyle to help with my burnout

The very next thing that I did was pray, and I felt God tell me "stop." I initially dismissed it, but I realized going at full speed hadn't gotten me anywhere. It was time to stop and smell the roses, and know that I gave it my best.

The first thing I did to get myself out of my rut was to go on walks, inviting my son along when he was ready. Then, I started cooking and cleaning more. I liked how it felt not being a robot whose only task was to apply to jobs.

The biggest moment of change was when my son and I started going back out again. It was back to the basics. Even if we don't have money, we're still going to have a good time. That's when I started easing back into myself.

I only apply to jobs for 90 minutes a day now

I picked up a part-time job in retail at Bath & Body Works to make some extra money. Additionally, I only give myself 90 minutes a day to apply for jobs, probably a few days a week.

That time usually includes me finding a job I want to apply for, using AI to optimize my résumé, submitting the application, and finding someone to follow up with. I find a lot of jobs from alum groups or referrals. If someone calls me back, great, but if they don't, it's OK. I can't afford to let the job market hurt my spirit anymore.

Outside of that, I follow a schedule. I wake up, take my son to school, work out, cook, and study the Bible or read another book. Right now, I'm challenging myself to read one audiobook and one physical book each week.

My discipline and spirituality are bringing me out of burnout

My spirituality has also kept me grounded during this time, and I keep myself very busy.

I'm currently looking for a hobby that'll keep me out of survival mode. I'm thinking of getting more into decorating because I get to use my hands, and it keeps my mind solving problems in a creative way. It's not easy, and it's not stressful.

I've been calling my loved ones more because I have the time, and I've been volunteering more. I've met good people there, and it makes me happy to serve.

One of the biggest things that's helping me ease out of burnout is having the discipline to stay busy and keep up with my schedule. No matter what I'm doing in a day, I treat it like it's my job.

Do you have a story to share about navigating burnout? If so, please reach out to this reporter at tmartinelli@businessinsider.com.

• AWS executive Sarah Cooper routinely crafts proposals to create new job roles and initiatives.
• Cooper uses Amazon's 6-page narrative memo and its PR/FAQ frameworks to pitch ideas.
• She wrote a proposal that led to her current role leading the AWS AI-Native team.

AWS executive Sarah Cooper has launched several companies and held roles at NASA and AWS, but she's only ever formally interviewed for three jobs.

Instead, Cooper told Business Insider that her career moves often begin with writing a proposal about the work she wants to pursue, either to a VC, a cofounder, or an executive leader, and that becomes the role she steps into.

"I have a million ideas," Cooper said. "It's a blessing and a curse."

Cooper has seen a high rate of success with her proposals. She used them to invent both her previous and current roles at AWS.

Cooper was previously the head of AWS Industry Products, where she worked closely with customers to develop industry solutions in the cloud. Now she leads the "AWS AI-Native" team, an AWS product organization that works with customers to understand how companies will evolve with AI.

Cooper said she wrote a proposal to pitch her last two roles, and she's a fan of the six-page memo, a favorite of Jeff Bezos. Instead of PowerPoint decks, Bezos often opted to start meetings by having attendees read a 6-page "narratively structured" memo silently at the start of the meeting. The founder has said he prefers memos to avoid "hiding a lot of sloppy thinking in bullet points."

These six-page memos typically describe detailed plans for a new initiative and are written using 10-point font. Bezos wrote years ago in an annual shareholder letter that "great memos are written and re-written," and can't be done in a couple of days, though the exact criteria remains more flexible.

Cooper also referenced using Amazon's classic PR/FAQ framework. The executive said she finds the format "effective at helping people quickly understand what your idea is and why it matters."

Amazon describes the framework as a means to develop ideas and initiatives that enable teams to quickly integrate feedback and promote fact-based decision-making. The press-release portion spans a few paragraphs and provides the reader with highlights of the customer experience. The FAQ section, typically five pages or less, dives into the details and also outlines the challenges of building the product or service.

Once a draft is written, the creator typically schedules a one-hour meeting with stakeholders to review the document and get feedback, the company said in a blog post about the framework.

Cooper said she's written over 14 PR/FAQs and investment narratives that have been funded, as well as many other ideas that the company didn't move forward with.

Cooper, who has been at Amazon for over nine years, says she also tries to reflect the organization's values when writing proposals. Amazon's PR/FAQ format originated from a process the company calls "Working Backwards," a method for evaluating new ideas by first defining the ideal customer experience and then reverse-engineering what the team needs to build to achieve it.

Cooper said Amazon is "customer-obsessed," so she tries to focus on that value when drafting a memo. She said she thinks less about the needs of a specific team and more broadly about customers and the best way to serve them, regardless of what's been done previously.

While Cooper has spent her career writing proposals, many never see the light of day — and those that do don't always go smoothly, she said. In her late 20s, while serving as the CTO of a company, she wrote a proposal suggesting that the business should focus limited resources on scaling the supply chain, specifically with the supply chain involved in on-mine extraction to meet demand for contracts. The company disagreed on where to invest.

"I got pushed out of one of my companies, basically, for a proposal I wrote," Cooper said.

• With the year coming to an end, gift-giving season is in full swing.
• Holidays can be crucial for maintaining the typical closet, from viral sneakers to a classic vest.
• Business Insider wants to know what employees in the finance world crave most this year.

While a big bonus or a long-awaited promotion might top Wall Street's holiday wish list, those can't exactly be found in stores or online.

Business Insider wants to know the more attainable gifts on bankers and traders wish lists this year — think the viral sneakers showing up on trading floors.

Whether it's something you're coveting yourself or a go-to item you're planning to buy in bulk for work friends, we'd love to hear it.

Your list might look different if you're a partner or an intern, and that's fine — no judgment. We just want a clear picture of what Wall Street really wants this holiday season.

Tell us more in the survey below.

• The drama over who will buy Warner Bros. Discovery overlooks viewers' shift toward YouTube and TikTok.
• Traditional studios face pressure as tech takes over entertainment and people prefer UGC platforms.
• Viewers can expect streaming prices to rise and theater releases to decline, regardless of who wins WBD.

The Warner Bros. Discovery-Paramount-Netflix drama has Hollywood in a tizzy this week — but it's missing one big thing: What about you?

Americans are deep in a new era of entertainment and are increasingly likely to get their fix on YouTube and TikTok rather than with WBD's popular franchises like Harry Potter and "Game of Thrones."

The headlines around the deal have focused on whether Netflix or Paramount's David Ellison will win in the boardroom. But the results in your living room will be due to forces that are much bigger than the companies themselves, and have changed the way we watch. Movie theater releases are down, cable TV is dying, streaming prices are likely to rise, and Hollywood's power is waning in the age of tech giants.

What the fight for Warner Bros. could mean for jobs and how you watch TV

The reason companies like Netflix and Paramount want to buy Warner Bros. is because competition in the entertainment industry is fierce, driven by the decline of cable TV and the rising popularity of small screens and tech companies.

"If Big Tech never made a move into media, there's a very good chance we wouldn't be seeing this much consolidation as we are now," said Brandon Katz, an entertainment analyst at Greenlight Analytics.

Now that Americans spend more time with the big players in entertainment like Google-owned YouTube, Netflix, and Amazon, mergers can help legacy players like Disney, Fox, and Comcast compete. It'll mean lost jobs. The combo that formed WBD led to multiple rounds of layoffs, including about 1,000 roles in mid-2024. Both Paramount and Netflix present themselves as the bigger job savers, but both promise billions in "synergies" that will likely result in cuts.

"The problem is the acquisition and pending consolidation of two media giants, not who the buyer is," said The Writers Guild of America, the union representing TV and film writers.

Additionally, deals offer companies the opportunity to consolidate their brands. WBD has already combined HBO's prestige shows with Discovery's mainstream fare to form HBO Max, and Disney has brought Hulu and ESPN+ into its streaming bundle.

It all trickles down to what you get to watch, where you watch it, and how much you have to pay. Streaming prices have been increasing while big theatrical movie releases have declined over time, and there's good reason to think those trends will continue, regardless of where Warner Bros. ends up. Industry trends for movie-going are clear, with the number of wide releases and box-office grosses still below their pre-COVID levels.

Streamer prices could increase while less stuff gets made

Whoever ends up with Warner's HBO Max will likely bundle it with their existing streamer and raise the price based on the more expansive offering, Katz said.

"Within 18 months, all consumers should expect to be paying more for these subscriptions," he said.

It's also widely expected that the combination of two of these companies will lead to less output. That's one of the main objections of Hollywood, whose livelihood depends on stuff getting made.

"This is one less entity writing checks," said Josh Rosenbaum, a producer at Waypoint Entertainment, whose credits have included Yorgos Lanthimos' Oscar-winning "The Favourite."

Paramount promised that if it wins the WBD battle, it'll release 30 movies a year in theaters, while Netflix says it'll continue to put Warner Bros. movies in theaters while favoring a quick-to-streaming strategy for its own films. The Disney and Fox merger is a cautionary tale; their combined theatrical releases have declined nearly 50% since the merger was announced in 2017.

The dinosaurs may be fighting a losing battle. People (especially younger generations) are increasingly getting their entertainment from user-generated platforms. A Hub Entertainment Research survey found that people aged 13 to 24 spend an estimated 21% of their screen time watching non-professional videos on platforms like YouTube and TikTok, compared to just 16% spent on traditional TV shows. In contrast, people aged 35 and up spend 39% of their screen time on TV shows — more than twice as much as they spend on online videos (14%).

That's put pressure on traditional players. YouTube — not any traditional Hollywood powerhouse — is the No. 1 streaming service on TVs, according to Nielsen. Netflix's engagement per user has stagnated over the past few years, despite the company's increased spending on content, while YouTube has widened its lead on TV screens, according to Bernstein research. Netflix and other leading streamers have taken steps to bring social media-first creators and video podcasters onto their platforms, but show no signs of fundamentally changing their models.

"The real threat to the media industry in general is the new formats like short form," Bernstein analyst Laurent Yoon said. Paramount is at least talking about a technology-led overhaul of how entertainment is made, which has investors intrigued. But whatever changes it makes are likely to be incremental in the near term, he said.

• Early menopause at 29 deeply impacted Claudia Zimmerman's personal life and work as a nurse manager.
• Medically induced menopause brought her years of anxiety, hot flashes, and weight gain challenges.
• Access to menopause-specific healthcare and support at work finally improved her well-being.

This as-told-to essay is based on a conversation with Claudia Zimmerman, 48, who went through medically induced early menopause at 29 while working full-time as a nurse manager. The following has been edited for length and clarity.

I've been a registered nurse for almost 25 years, and I've spent the bulk of my career working in labor and delivery and at a fertility clinic. I always felt passionate about women's reproductive health due to my own experiences with endometriosis since I was a teen.

I had stage three endometriosis, a disease that happens when tissue grows outside the uterus. The pain was excruciating, and I went to multiple physicians who dismissed my symptoms as something women just go through.

This was frustrating and isolating, and it made me doubt the symptoms my body was showing me.

I had to do something to help with the pain

At 29, I was done having children, and my symptoms were escalating, so I decided to have a hysterectomy to relieve my pain once and for all. The hysterectomy removed my ovaries, uterus, and cervix, which put me into medically induced early menopause.

I knew of the general symptoms of menopause, but nothing could've prepared me for the experience I was about to undergo.

The next few years were a struggle between my body and menopause. One of the most challenging aspects was working full-time while going through it.

I was working full-time as a nurse manager at a fertility clinic

After my hysterectomy, I went back to work as an IVF nurse. I didn't become menopausal right away. I had an ovarian remnant that was producing enough estrogen to hold off menopause.

Two years after my hysterectomy, they removed this third-ish ovary, which then led to the first wave of my menopausal symptoms. I experienced increased anxiety, a bit of weight gain, and increased depression in addition to hot flashes, which I mitigated with estrogen tablets and patches.

I powered through hot flashes, even though they took a toll on me

My hot flashes were a slow build that would start with a sudden panic. I'd feel pressure in my abdomen and chest that would take my breath away.

I'd keep getting warmer and warmer to the point where my cheeks would get red, and sweat would leak out all over my body.

It didn't matter if it was snowing outside or freezing inside; for 30 seconds to up to three minutes, I'd be drenched in sweat, and my heart would race.

I was open with my coworkers about what I was going through, and that was helpful

I worked alongside many women. I would often turn to them for advice on my symptoms, and they helped me feel empowered and supported.

At work, I tried a menagerie of things to help combat my symptoms, from drinking iced water, eating straight ice, and bringing in my own fan for my desk, to taking a walk on a break while talking to my mom, who always helped me work through anxiety and fear.

I was so self-conscious about the weight gain. I started wearing larger scrubs so that I could conceal some of the changes, but there were also days when wearing scrubs wasn't an option, like when meeting new patients or traveling from office to office.

Given that my job was very patient-centric, back-to-back interactions dominated most of my day. On days when I wasn't feeling great, I knew I could share with my colleagues that I "needed a minute," and they would step in immediately to offer their support.

A lot of my coworkers and I were the same age, but our bodies were different

My coworkers and I were all in our mid-30s, yet our bodies were so different. Although they never made me feel that difference, it was hard for me to ignore.

I often had to remind myself of why I made the decision to get a hysterectomy. It was saving me from decades of severe pain, even though it was forcing me to come to terms with a new version of myself I wasn't ready for.

I was most shocked at how long my menopause journey lasted

I experienced menopause symptoms for roughly 17 years. The weight gain and hot flashes were one thing, but the anxiety continued to escalate. I was not prepared for how it seemed to overtake my life seemingly overnight.

All these changes, coupled with the pressure of my role as a mother and wife, left me feeling resentful and short-tempered. It was hard to sleep more than a few hours at night.

I tried to mask my symptoms as much as possible

I was so uncomfortable all the time because of these changes. I didn't want others to see me as weak or different, or to judge me.

Additionally, I was caring for people every day who were going through journeys far worse than my own. I had three children, and I was taking care of people who would do anything just for the ability to have one. I wanted to be strong for them.

After 10 years in this fertility practice, I transitioned into benefits management with Progyny, a fertility, family building, and women's health benefits company, as a clinical educator and patient care advocate. When I started working at this new job, just after I turned 40, everything began shifting dramatically.

My menopause symptoms reached their peak

In 2018, my body played every trick known to man on me. I was averaging about three to four hours of broken sleep a night. My diet didn't change, but I felt like I had gained 10-15 pounds overnight.

My anxiety spiked even further. I felt the brunt of it in the evening and the night.

The weight of fear, worry, guilt, and shame would overcome me, and my thoughts would race during the hours I was supposed to be relaxing and resting. To top this off, my libido vanished, and the short temper I mentioned previously turned into straight anger.

I also experienced a crippling loss. My sweet mother, who had been diagnosed with ovarian cancer only a year before, succumbed to her disease. This devastation wrecked any stability I had left of my mental state.

My job had benefits that were very helpful

I was now working from home, so I no longer had to commute up to three hours round-trip to work or be in the clinic.

This also meant I didn't have to mask my symptoms from colleagues and patients, but it began affecting my family and loved ones. My "mask" that I could keep while I was a nurse fully came off.

To distract myself and shield my husband and children from the person I had become, I threw myself into work. It was the only place I felt any sense of control. My family knew how much I was suffering, and my husband was very helpful.

My new role focused on helping companies offer fertility, family building, and women's health benefits to their employees. Through my work there, I stumbled on a breakthrough.

My insurance offered me specific coverage

At my new job, I was helping develop a menopause-specific benefit. I was both shocked and elated to learn that menopause-certified clinicians were available through a new digital health platform.

I thought about how incredible it would be for women to access this type of coverage, regardless of how their menopause stories unfolded. As soon as the product went live, I made my first appointment with a menopause specialist.

I cannot describe how life-changing that appointment was for me

The nurse practitioner listened intently as I described my symptoms, as well as the misery and isolation I had felt several years prior. Every aspect of my health was carefully considered, and I received proper, focused care.

After three new medications, I was able to sleep eight hours again. I lost 25 pounds, all while maintaining balanced hormones and a balanced diet. I still have anxiety from time to time, but it no longer weighs me down. I'll spare you the details, but my libido once again exists.

The most important thing an employer can do for its female population is to acknowledge the unique healthcare challenges that women face and be proactive by offering access to this specific type of care. That, and an empathetic ear, will go a long way for the women of their organizations.

Menopause can feel like you're being introduced to somebody that you do not know, which can upend your life, in and out of work. I wish I had known at 29 to advocate for myself, and that eventually things would get better.

• Surge AI CEO Edwin Chen said data labeling is much more creative than it gets credit for.
• "I think a lot about what we're doing as a lot more like raising a child," Chen recently told "Lenny's Podcast."
• Surge AI is one of the larger startups competing to help train leading AI models.

It could be easy to dismiss the work data-labeling firms do. Surge AI CEO Edwin Chen said that could stem from misunderstanding what they do.

"I think a lot of people think of data labeling as it relates to simplistic work, like labeling cat photos and drawing boundary marks around cars," Chen told Lenny Rachitsky on his "Lenny podcast."

Chen, who previously worked at Google, Twitter, and Meta, said that he's "always hated the word data labeling."

"Because it just paints this very simplistic picture when I think what we're doing is completely different," he said.

Surge AI, which Chen founded in 2020, competes in the AI data labeling space with companies like Scale AI and Mercor. Surge also has a partnership with Anthropic and also runs DataAnnotation.tech, where freelancers can sign up to get paid for training AI models. These remote workers are often referred to as "ghost workers" for their behind-the-scenes labor that is critical to AI's development.

Beyond any rote work it can entail, Chen said data labeling is a much more creative endeavor. He compared what companies like Surge do to how parents instill lifelong values in their children.

"I think a lot about what we're doing as a lot more like raising a child," he said. "You don't just feed a child information. You're teaching them values, and creativity, and what's beautiful, and these infinite subtle things about what makes somebody a good person."

In this way, Chen said, companies like Surge AI are "raising humanity's children."

Chen's view can also be seen when you navigate to Surge's website, which greets visitors with the question: "What made Hemingway, Kahlo, and von Neumann extraordinary?"

"Their life experiences: war, love, triumph, loss. The people they met, the cities they explored, the thousand choices that made them who they were," the website reads. "Data does for AI what life did for them — transforming it into intelligence that could one day prove the Riemann hypothesis, imagine new philosophies, and send rockets to the stars."

'You don't need to become someone you're not'

Chen previously worked in Big Tech at companies including Twitter, Google, and Facebook — and you might remember one of his Twitter projects.

While at the company, roughly a decade before Elon Musk would acquire it, Chen became known for making the "Pop vs. Soda" map by geo-tagging data from tweets around the US to illustrate which word users used to refer to soft drinks.

Looking back at starting Surge AI, Chen said he was surprised to find out that he never had to stop burying his head in the data.

"I thought if I started a company, I'd have to become a business person looking at financials all day, and being in meetings all day, and doing all this stuff that sounded incredibly boring, and I always hated," he told Rachitsky. "So, I think it's crazy that didn't end up being true at all."

Chen said he wishes he had known that you don't need to spend "constantly tweeting and hyping and fundraising."

"You don't need to become someone you're not," he said. "You can actually build a successful company by simply building something so good that it cuts through all that noise. And I think if I had known this was possible, I would've started even sooner."

Do you work in data labeling? Contact the reporter from a non-work email and device at bgriffiths@businessinsider.com.

• Modular aims to break Nvidia's dominance in AI chips with a new portable software stack.
• The startup's platform lets AI models run on different GPUs, not just Nvidia's.
• Modular CEO Chris Lattner helped bring Google TPUs to market, with cofounder Tim Davis.

In Silicon Valley, where bold technical bets abound, few bets look bolder than trying to break the grip of Nvidia's CUDA, a software stack that's quietly become the operating system of the AI boom.

That's what Modular, a startup founded by software gurus from Apple and Google, is trying to do.

Cofounders Chris Lattner and Tim Davis have spent decades building the software plumbing that sits beneath the modern tech industry. Lattner is famous for creating Apple's Swift programming language. He also built the software underpinning Google's TPU AI chips, with Modular cofounder Tim Davis.

They're now aiming that expertise at CUDA itself. The attempt borders on madness, but it's the kind of audacious project that could transform the AI industry.

"It's seen by a lot of people as somewhat crazy," said Kylan Gibbs, CEO of startup Inworld AI and a former product manager at Google DeepMind. "That's where Chris has the advantage: He's smart enough to actually know how to do it, and somewhat crazy enough to set out to do it."

CUDA entrenched. Competition fragmented.

CUDA began life almost 20 years ago as a way to make graphics chips programmable. Today, it has grown into a multilayered software ecosystem — language, libraries, compilers, inference systems — that most AI companies rely on.

That success comes at a cost: Most of the industry is now optimized around a single vendor's hardware. CUDA binds AI workloads to Nvidia GPUs. That is great for Nvidia, but deeply limiting for everyone else.

On the surface, there seems to be a ton of competition: AMD sells GPUs. Google has TPUs. Amazon created Trainium AI chips, and a host of startups are building similar hardware.

The problem is that each chip comes with its own software stack optimized just for that component. That means an endless reinvention of the wheel. Most of the time, it's simpler to just stick with CUDA — and Nvidia's GPUs.

And yet, AI developers crave portability: Being able to use any combination of GPUs from multiple providers without juggling different software stacks.

"Nobody is building portable stuff, because why would anyone work on software for more than one chip when the chip projects themselves are doing the software?" Lattner, Modular's CEO, told me in an interview.

Nvidia could extend CUDA to run well on rival AI chips. But doing so would undermine Nvidia's greatest moat: the closed-loop bond between its software and its chips. "Obviously, they don't want portability," he said.

Paradox = opportunity

For Lattner, this paradox presents a big opportunity.

"We realized there's nobody in the industry that's actually incentivized to do this. It's very expensive, very hard," he said. "And at the same time, everybody wants it."

That's what inspired Lattner and Davis to leave Google and start Modular in 2022, the year ChatGPT took the world by storm.

Since then, Modular has raised $380 million from investors including Greylock, General Catalyst, and GV, Google's venture capital arm. The latest financing in September valued the startup at $1.6 billion. Modular isn't the only effort to break the CUDA lock-in. There has been ZLUDA, an open-source project that was funded by AMD, and more recently, the startup Spectral Compute, which has raised $6 million.

Lattner has used some of this money to hire talented programmers from Google, Apple, and other tech companies. They spent three years working in relative obscurity to create the building blocks of a new AI software stack.

The new AI software stack

The foundation starts with a brand-new programming language, called Mojo, that offers deep controls for making AI chips run as efficiently as possible.

Modular designed this to work similarly to Python, a popular and easy-to-use programming language. But Mojo also has the speed and power of other, more complex languages, such as C++, that are essential for AI development. Mojo also works well with PyTorch, an open-source framework that is often used when building AI models and applications.

I first heard about Modular earlier this year when interviewing Carles Gelada, a former OpenAI researcher. "There are several interesting projects to create GPU-agnostic frameworks and platforms, and challenge CUDA," he said at the time. "Mojo is the most interesting one."

MAX is the next major layer of Modular's new software stack. It powers AI inference, which is how models are run. This part of the system works with Nvidia GPUs, AMD GPUs, and similar chips from Apple. Modular hopes to add more AI chips in the future.

On top of that is another layer called Mammoth, which helps AI developers manage GPU clusters.

In late September, Modular announced that it got top performance out of Nvidia's new Blackwell B200 GPUs and AMD's latest MI355X GPUs — crucially on the same software platform.

Lattner said Modular got these AMD GPUs to perform roughly 50% better than when these chips run on AMD's own software.

More importantly, the ability to run different GPUs on the same software stack now supports the tantalizing opportunity to compare Nvidia's offerings with rival AI chips on a more level playing field.

"The obvious question is: can MI355X compete with Blackwell?" Modular wrote in a blog announcing the results. "Early signs point to yes."

A customer test

Gibbs, the CEO of Inworld AI, has been putting Modular's software through its paces in the real world.

Inworld builds high-speed, real-time conversational AI technology that supports offerings from big companies, including Disney, NBCUniversal, and Niantic Labs.

Earlier this year, when the startup designed a new text-to-speech AI model and got early access to Nvidia B200 GPUs, they issued Modular a challenge: Cut our costs by 60% and reduce our latency by 40% and we'll work with you.

"Within about four weeks, we were able to get this incredible performance," said Gibbs, who signed a partnership deal with Modular soon after. "I've bet with my wallet."

While Inworld was mostly lured by Modular's performance gains on Nvidia's latest GPUs, Gibbs likes the flexibility of using different AI chips more easily in the future, if needed.

"The promise is that we'd be able to move to new hardware," he said. "Let's say AMD takes off, let's say TPUs take off for Google, or there could be other new hardware that comes online. So it's nice to have that optionality."

'Nvidia doesn't have to die'

In fact, Google's TPUs are suddenly having a moment. The internet giant released a new AI model called Gemini 3 to rave reviews recently. That was trained and run using TPUs, and some other AI companies have signed deals to use these chips instead of, or alongside of, Nvidia GPUs.

That's put Nvidia on the defensive. A project like Modular, with its promise of portability across different AI hardware, adds to this pressure.

"Nvidia could kill this in a day," said Gibbs of the Modular project. "Nvidia could basically say, 'okay, we don't really care that you run just on Nvidia hardware. Here's a CUDA option that runs on AMD GPUs as well.' It'd be a bit crazy for them to do that, but it's something they could do and that would of course be somewhat bad."

For all Lattner's critique of the industry, he says Modular is not trying to kill Nvidia. In fact, he argues that Nvidia will continue to thrive, even if Modular succeeds spectacularly.

"We're trying to build something like Android, but for AI hardware," he told me, referring to Google's mobile operating system that powers most of the world's smartphones.

Despite billions of people using Android devices, this success didn't kill iOS, Apple's mobile operating system. iPhones still rule in the US, for example.

Lattner thinks something similar will happen in AI as Modular's software makes other hardware more competitive, giving developers more freedom, and chipping away at the industry's single-vendor monoculture.

"So Nvidia doesn't have to die, but we do want more competition. We do want more innovation," he said. "I think that's good for the world."

Sign up for BI's Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com.

• EQT is the world's second-largest private equity firm, but it's not a household name in the US.
• Peter Aliprantis, EQT's US private wealth head, aims to change that and attract affluent investors.
• He walked BI through the firm's plans to win over retail investors and why it's grown so much.

EQT is one of the largest private equity investors in the world — yet most wealthy Americans have barely heard of it. That's the uphill battle facing Peter Aliprantis, the Swedish firm's head of private wealth in the Americas, as EQT tries to pitch in a market dominated by Wall Street brands with plenty of CNBC airtime.

"Most people in the United States are not familiar with us, and the way we say it, we're the best-kept secret," Aliprantis told Business Insider.

Private Equity International ranked the firm as the second-largest private equity firm, with $312 billion of assets under management. It raised more than $113 billion in third-party private equity capital from 2020 to the end of 2024, putting it ahead of Blackstone, and just behind KKR so far this decade.

Like many of its competitors, it's turning to private wealth as the newest source of growth. The industry's change of fundraising focus comes as private equity firms are slow to return cash to investors, and over-allocation among institutional investors means that institutional funding is slowing.

But the same reasons that the firm isn't as well-known in America are actually an advantage, Aliprantis said.

In a world where debt-heavy buyouts are proving more difficult and an increasingly concentrated American private market is pushing some to invest internationally, a global industrialist approach can be attractive.

EQT has returned capital at a normal pace, with $23 billion in distributions for the year ending June 2025. The firm has also been building a private wealth business for the past four years, which accounts for 10% of its current assets. The firm has a goal to reach between 15-20% during its current $100 billion fundraising cycle, according to its second-quarter report.

Aliprantis walked Business Insider through the firm's pitch to financial advisors and private wealth distribution networks, explaining why its global reach is a significant advantage in 2025.

The key for EQT, Aliprantis said, is for the firm to offer individual investors the "exact same deals" it gives institutional investors.

EQT's industrialist, international advantage

EQT was founded in 1994 as a spin-off from industrial holding company Investor AB, but the firm's history stretches back to Sweden's Wallenberg family. The Wallenbergs, called the "Rockefellers of Europe," have created an empire of business holdings including massive stakes in Sweden's biggest firms, like ABB, AstraZeneca, or Saab.

"The Wallenberg family has a 160-year heritage of owning and developing companies," Aliprantis said. "We're not financial engineers. We don't add a lot of leverage to what we do, and we're very, very different from what a lot of our peers on Wall Street are doing."

Aliprantis's comments echo a larger change in the industry, which is running out of easy money-making deals and cheap financing and now has to extract returns by actually building stronger companies.

But the firm's biggest advantage, Aliprantis said, is its global nature.

Only 35% of its assets are based in North America, and the firm has 26 global offices where its deal teams invest in local private equity, infrastructure, and real estate deals.

"A lot of our colleagues based in New York will fly deal team partners over to different places around the world to do the deal and then get on a plan and fly home," Aliprantis said. "Our deal teams are pretty much based in the locations where they do deals."

This means the firm "gets the call" when local companies are looking to sell, and keeps them from larger "bake-offs" where the price might be bid-up.

This has also meant the firm can continue to provide distributions to its clients even if the market is slow in one locale.

"If you're a US-based domicile private markets firm that has 70 to 80% of your assets in the US, guess what? If the US IPO market is slowing, you're going to have a problem exiting," Aliprantis said.

"Here in the US, it's always been too much money chasing too few deals. You know what? That's a US thing," Aliprantis said." If you go to Europe and you go to Asia, it's the opposite."

For example, Bain estimates there's about $480 billion in dry powder for European private funds, including venture capital, compared to Pitchbook's $914.5 billion for US-focused private equity firms, not including VC. Apollo's Marc Rowan also recently told the Wall Street Journal that as an industry, they find themselves short ideas rather than capital.

Aliprantis said investors' biggest reason to diversify away from the US market is its concentrated bet on AI.

"Their concern is that the Mag Seven is roughly 37% of the S&P right now, and valuations are stretched," Aliprantis said. "Is AI really going to work? Is it not? How additive is it going to be to the bottom line? We don't know."

How to keep retail investors happy

Across the spectrum, Aliprantis said, the "biggest concern" is that retail investors are getting a set of less attractive deals, while institutional investors are getting a "separate set of deals."

Aliprantis said that the firm's six evergreen vehicles are composed of the "exact same deals" that its institutional clients invest in.

The key to doing that, and to being a responsible investor or retail capital, is "size and scale," Aliprantis said.

Size also helps with the balance sheet necessary to launch a private wealth business. It can cost millions of dollars to hire the necessary staff to start selling to financial advisors and other wealth management channels before any revenue is returned to investors.

EQT was able to use its balance sheet, as a public company in Sweden, to build its private wealth team and now has 70 private wealth professionals globally, with 20 based in the US.

That's not to say that smaller funds won't succeed, but it will be much harder, Aliprantis said. With so many investors competing for retail capital, consolidation is inevitable.

"The race is on in the industry right now," Aliprantis said.

This summer, after Meta made a $14 billion investment in Scale AI and poached its 28-year-old founder, Alexandr Wang, and after A-list clients like OpenAI and Google halted work with the startup, a worker for Scale AI anxiously asked ChatGPT what it thought of his company's fate. He knew the chatbot well, having tested it for vulnerabilities. Its prognosis was grim.

"Scale AI will no longer exist as a credible independent entity within 24 months," ChatGPT, which isn't any kind of official authority, wrote. "Its infrastructure will be repurposed for Meta's internal needs. Its client base will evaporate. Its role as a neutral red teamer or external evaluator is effectively over."

The contractor shared the chat logs with fellow workers at Scale AI. In one reply reviewed by Business Insider, one worker said that they were already on their way out, describing the startup as a ticking time bomb.

The chatbot was drawing from a storm of headlines about Scale AI, which until this summer had been touted as one of the most ascendant startups in tech — the place Big Tech companies vying for AI supremacy went to when they wanted their chatbots stress-tested and perfected. Lately, it's lost some of its gleam, with investors significantly lowering valuations, workers sniping about pay, and rivals coming for its clientele.

The vast army of human data labelers that made Scale AI a juggernaut are chafing at what they say are pay cuts, lengthy unpaid onboarding sessions to join new AI projects, and thinning workloads — and are increasingly leaving the platform altogether, according to interviews with five current and former contractors and internal correspondence obtained by Business Insider.

Activity in the main internal chatroom for Outlier — Scale AI's flagship gig work platform, which touts more than 100,000 taskers — has plummeted since the Meta investment, with weekly discussion threads drawing dozens instead of the usual hundreds of replies, according to screenshots reviewed by Business Insider.

One tasker said that they'd spent close to 40 hours in a single month in unpaid onboarding sessions without landing any actual work, noting that other platforms like Scale AI's rival Mercor do pay for this kind of work. Elizabeth Boyd, another tasker, says she rarely does work for Outlier anymore after seeing effective pay rates for some projects slashed to around $20 an hour — down from the $50 she used to make. One gig that advertised $20 an hour only allowed three minutes of working time every two days, or a 99-cent payout, according to screenshots obtained by Business Insider.

Joe Osborne, a Scale AI spokesperson, says the balance sheets show the company is on the right path. "This quarter is on track to be our biggest of 2025, our data business is more profitable today than it was before the Meta deal, and our applications business, which includes work with Fortune 500 companies and governments, has doubled revenue" in the second half of the year compared to the first, Osborne wrote in an email. He also noted there has been an increase in active users on Outlier since the Meta deal, and that pay rates are based on the skills for each project and contributors always see the rates upfront, and have the option of declining any gig.

The company is also looking to diversify. The startup has embraced fields like robotics, announcing a new lab to meet booming demand for robot training data this fall. It's doubling down on its US military and other government work, winning up to $199 million worth of defense contracts since the Meta deal.

Some investors are bullish. In one current investor's view, Meta has mostly left Scale alone, letting it operate as an independent company. With around $1 billion on the balance sheet, the investor added, there are no plans to fundraise. And an IPO could still be on the table at some point.

Other investors see Scale AI as more like a gutted fish. The Meta investment — which valued Scale AI at $29 billion — has dented Scale AI's valuation in private markets where people buy and sell equity from pre-IPO startups. Noel Moldvai, Augment's CEO, tells Business Insider his platform used to process millions of dollars' worth of transactions in Scale AI stock before the Meta deal, but that dried up as sellers waited to see if the startup rebounded. Activity is picking up again, he said, but at lower valuations of around $15 billion to $9 billion. The underlying message of Meta's semi-acquisition is clear to Moldvai. "It seems like Meta was just after Alexandr Wang, and so this is probably the structure that let them get him," he says. He adds that Scale AI's valuation could still bounce back.

On another marketplace, Caplight, Scale's valuation has dropped to $7.3 billion. Osborne says that the valuation is not accurate because there have been no sales of stock at that price and multiples of comparable companies would yield a higher valuation.

If the company doesn't pull it off, it could become the latest example of a once-promising startup that morphed into a "zombie" after being invested in by a tech giant.

---

This summer, Scale AI painted a rosy picture of the Meta investment, describing it as a major cash infusion and source of future work in its official statement at the time of the deal. That messaging was also conveyed to the startup's corporate workforce, says another former member of Scale AI's red team, which tests chatbots for flaws and vulnerabilities. Just a few weeks after the investment, he was laid off as part of a major downsizing that saw 14% of Scale AI's full-time staff of 1,400 let go.

Osborne said the layoffs were aimed at making the data division profitable, which it now is.

Those weren't the only cuts. In September, Scale AI terminated 12 contractors on its red team, citing performance issues. Two ex-red teamers told Business Insider that the team's work had been drying up since the Meta deal, blaming thinning workloads for the cuts. Later that month, Scale AI shuttered a team in Dallas of contractors focused on generalist AI work as it moved towards more specialized fields.

Osborne said the 12 contractors were part of Scale's temporary workforce and represented a small fraction of its overall red team, which the company is still committed to investing in. He said the Dallas cuts were part of an industry shift towards higher skilled work and represent a small fraction of its overall workforce.

Meanwhile, a swarm of AI training startups has rushed in to poach Scale AI's workers and clients. Some are now raising capital at soaring valuations. Surge AI hit a valuation of $24 billion while Mercor, which is run by three 22-year-olds, announced in October it had raised $350 million at a $10 billion valuation.

Mercor has won at least one major AI training project from Meta, Scale AI's 49% shareholder. In September, Scale AI filed a lawsuit in California against Mercor alleging it hired one of its sales employees to poach its biggest customers, allegations Mercor denies.

Spam and low quality data became accepted as a cost of doing business.Tammy Hartline, a former Scale AI consultant who now works at Mercor.

One Scale AI investor said he'd been frustrated with Scale AI's leadership losing customers to Surge AI in particular, which reportedly brought in more revenue than Scale in 2024, despite never having raised outside funding. (Scale AI had raised more than $1.5 billion before the Meta deal.)

Brendan Foody, Mercor's CEO, has publicly criticized Scale AI for what he claims are low pay rates and data quality issues. "Scale lost the focus on product, on scaling quality," Foody said in a September podcast appearance. In response, a Scale AI spokesperson told Business Insider its quality metrics are at "record highs."

It's not just rival CEOs making that point. Tammy Hartline, who managed projects for Scale AI as a consultant until the summer of 2025, said Scale grew so quickly that the work became more about needing bodies than skills. "Spam and low quality data became accepted as a cost of doing business," she said. Hartline joined Mercor in September.

Scale AI has also been beset with security issues that predate Meta's investment. In June, Business Insider reported that Scale AI routinely used public Google Docs to track work for high-profile customers, including Google, Meta, and xAI. That practice left AI training documents labeled "confidential" accessible to anyone with the link and exposed reams of personal information, like private emails and pay details, about contractors. "We take data security seriously," Osborne said. "We conducted a thorough investigation and disabled any user's ability to publicly share documents from Scale managed systems."

Sloppy security isn't uncommon in the AI training space — Surge AI similarly left open sensitive work for its client Anthropic. But the exposed documents seen by Business Insider show that for a project for Google, Scale AI faced security and quality issues throughout 2023 and 2024. Thousands of taskers were flagged for being "suspected spammers," "cheaters," with hundreds of workers listed in spreadsheets with titles like "Good and Bad Folks" and "suspicious non-US taskers." Meta recently removed more than 40 groups buying and reselling AI training accounts, including from Scale AI, in response to a recent Business Insider investigation. Osborne said Scale's data quality metrics are the highest they've ever been.

The company has notched some recent wins. It once appeared bogged down by litigation, but recently agreed to settle multiple lawsuits filed by ex-workers in California who alleged they were underpaid and misclassified as contractors. (Scale no longer accepts gig workers from the state.) The big question remains whether Scale AI can thrive in the increasingly competitive AI training industry it helped give birth to. For many former workers, it'll be too late to find out.

---

Charles Rollet is Business Insider's tech correspondent in San Francisco. Ben Bergman is a senior correspondent at Business Insider, where he investigates the tech industry with a focus on venture capital and startups.

Last month, "Walk My Walk" hit number one on Billboard's Country Digital Song Sales chart. The moody stomp-clap tune, with lyrics like "Every scar's a story that I survived, I've been through hell, but I'm still alive," has been played more than 8 million times on Spotify. The song wasn't performed by a human artist — Breaking Rust, despite having the face of a handsome, rugged man in a cowboy hat on its Spotify profile, is an AI project. But there is a real person claiming that Breaking Rust's work is a copycat: Blanco Brown, an artist who mixes country and rap together, who claims the song's creator used AI to emulate his style. The person behind Breaking Rust did not respond to a message I sent asking about the origin of the sound.

It's the latest example of ways that AI-generated music, with its opaque origins, can create confusion around who really made a song just as easily as it can create a hit. AI-generated music that sounds a lot like your faves but was made with a few prompts has been going viral, spreading far and wide and more quickly than music labels can always have it removed.

When some of the first AI-generated tracks started racking up listens two years ago, music labels went to battle, threatening and filing legal action to stop AI generators from training on and using their artists' voices and music stylings. Universal Music Group (UMG) pushed to have a YouTube video where Eminem's voice rapped about cats taken down. Spotify removed AI slop songs that were listened to by bots to reap the streaming earning pool, and UMG also got streaming platforms to remove a viral "Drake" song that wasn't by Drake and The Weeknd at all, but a song written by Ghostwriter, an anonymous artist that uses AI to produce music and appears publicly only when cloaked in white and dark glasses.

Now, the labels are starting to drop their fists and shake hands with AI music generators.

Warner Music Group announced last month it had settled a lawsuit against AI music generator Suno (a test of the service by plaintiffs in the suit found that Suno would churn out works similar to ABBA and Chuck Berry when prompted in their style) and entered into a partnership with the company. Robert Kyncl, CEO of WMG, called it "a victory for the creative community that benefits everyone." The announcement came just weeks after UMG, the world's largest record label, settled its copyright infringement lawsuit with AI music generator Udio, and said the two have partnered to create a new subscription service, slated to launch next year, run on gen AI and licensed music from the label's artists.

Every minute that is spent listening to a generative AI track is a minute less spent listening to an artist track.Mark Mulligan, founder and senior music analyst at research firm MIDiA

AI companies face a litany of lawsuits after using copyrighted material to train their models. The battle is playing out in Hollywood, the news industry, and in visual arts — the music industry is the latest to decide it might be better to play nice with AI than continue a prickly, drawn out court battle when their rights sit in a gray area. "AI is here to stay, it's transformative," Chris Wares, assistant chair of the Music Business Department at Berklee College of Music. The record labels, he says, are "futureproofing themselves."

The proliferation of AI-generated voices and music stylings has sent a flood of new songs — some good, many uncatchy slop — onto streaming platforms and social media. There are more than 100 million songs on Spotify, Apple Music, and Bandcamp, and many are rarely or never played. Deezer, a French streaming platform, said in April that people were uploading some 20,000 fully AI-generated tracks each day, comprising nearly a fifth of all new content. As more playlists and artists making gen AI music are uploaded, human artists must fight for your ears. In July, a group called The Velvet Sundown rapidly racked up 1 million listens on two albums, something that many indie bands struggle to do on streaming platforms — but the photos of the band on social media are AI generated, and the real person or people behind the project remain unknown. In November, Billboard identified at least six AI or AI-assisted songs that had climbed onto its various charts. Amid the deluge, Spotify updated its impersonation policy in September, saying it would remove songs that featured the unauthorized use of someone's voice.

If someone makes a banger AI cover song or viral mashup, like reuniting Stevie Nicks and Lindsey Buckingham for a new Fleetwood Mac album or squashing the beef between Kendrick and Drake in a generated collaboration, the novelty factor could drive listening numbers that cut into original work from those artists. But the aim of these partnerships is to create new revenue streams for artists. The Warner deal stipulates that Suno will allow only paid accounts to download generated audio, and says artists must opt-in to having their names, voices, compositions, and likeness regenerated. In an industry where streaming has dramatically reduced royalties, that could be a boost — if fans make music of their faves, it could lead to passive income for the musicians. But that also means the artists' original works will be competing for ears against derivatives of themselves not just for your ears, but for streaming dollars.

"The reason why no generative AI music can be artist-first is because we are in a finite attention economy. Every minute that is spent listening to a generative AI track is a minute less spent listening to an artist track," Mark Mulligan, founder and senior music analyst at research firm MIDiA says. "We are definitely in a world now where more and more consumers are creating, and that is competing with entertainment time."

Part of that creative process may draw us back to older roots in how people interact with music. For hundreds, if not thousands, of years, music was communal. People played it together and used it to pass down stories. With recording technology and radios, music became widely distributed, which "created this moat between artist and fan," Mulligan says. "We got to this idea that music is a creative full stop, and that the audience doesn't help shape what the music is apart from when you go and see the band play live." But now, AI tools are becoming the ultimate form of fan expression. "We're widening the funnel of creativity," says Mulligan.

That's if artists authorize their voices to be used by the platforms. Some forward-thinking musicians, like Grimes, have already made clones of their voices and invited listeners to experiment. It's less clear if your typical popstar will OK the use of their voice to sing words they haven't seen, and if the potential new revenue streams of such an experiment would prove worth the risk. On Friday, Brown released a "trailertrap" remix of "Walk My Walk," a sort of taking back of his own style after seeing it emulated. So far, it has just 2,000 streams. "If someone is going to sing like me, it should be me," Brown told the Associated Press last month. Going forward, it will likely take more than one listen to know if the music we hear is performed by the artists we've come to love.

---

Amanda Hoover is a senior correspondent at Business Insider covering the tech industry. She writes about the biggest tech companies and trends.