• A version of this story originally appeared in the BI Tech Memo newsletter.
• Sign up for the weekly Tech Memo newsletter here.

Silicon Valley is sending a clear message about how staff should perform, and how it wants to pay them.

Instead of obsessing over punishing the bottom of the stack, Big Tech is increasingly choosing to shower rewards on the top:

• Meta’s overhauled performance system is the latest and most striking example. Under its new “Checkpoint” program, standout employees can earn bonuses worth up to 300% of their target. That’s tens of thousands of extra dollars tied purely to impact, not promotions or title changes.
• Google is loosening the funnel into its top performance buckets and shifting more bonus and equity upward.
• Amazon is doubling down on sustained excellence, allowing long-term top performers to earn above traditional pay-band caps.

What’s especially notable is how this is being driven by AI, according to Zuhayeer Musa, cofounder of Silicon Valley compensation data cruncher Levels.fyi.

As AI tools get more powerful, the value of high-leverage individuals is rising. The most prized employees today are “player-coaches,” people who ship work themselves while also guiding projects, mentoring peers, and shaping strategy. AI makes that hybrid role even more potent, allowing a single strong contributor to amplify their output, and their influence, without managing a large team, Musa wrote this week.

The result is a revival of the individual contributor track. You no longer have to become a full-time people manager to earn top-tier pay. At companies such as Meta, Google, Amazon, and Nvidia, senior ICs can now match (or beat) the compensation of managers simply by delivering outsized results, he noted.

For ambitious employees, the takeaway is simple: in the AI era, execution compounds. And Silicon Valley is increasingly willing to pay for it.

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

• OpenAI is officially preparing to test ads in ChatGPT.
• The move comes as the AI company looks to increase revenue amid $1.4 trillion in spending commitments in the coming years.
• CEO Sam Altman once portrayed ads as “a last resort” business model for the company before softening his stance.

Netflix famously backtracked on its stance toward ads. Now, OpenAI is following suit.

The AI pioneer announced that ads are coming to ChatGPT — less than two years after OpenAI CEO Sam Altman portrayed them as “a last resort.”

“Ads plus AI is sort of uniquely unsettling to me,” Altman said during an event at Harvard University in May 2024. “I kind of think of ads as a last resort for us for a business model.”

Altman’s softened stance since then underlines the massive change OpenAI has undergone in the last two years, and the company’s embrace of advertising is a testament to just how expensive the AI race has become.

In June, the OpenAI CEO said he wasn’t “totally against” ads, he just wanted to make sure OpenAI got the balance correct.

“We haven’t done any advertising product yet. I kind of…I mean, I’m not totally against it,” Altman said on OpenAI’s podcast. “I can point to areas where I like ads. I think ads on Instagram, kinda cool. I bought a bunch of stuff from them. But I am, like, I think it’d be very hard to — I mean, take a lot of care to get right.”

In October, Altman expressed a desire to make sure the company went about ads in the proper manner when asked about OpenAI’s past criticisms that other tech companies made addictive products.

“We’re definitely worried about this,” Altman said in response to a question that expressed concern about the similarities of Sora, OpenAI’s AI video app, and TikTok and the potential of ads. “I worry about it, not just for things like Sora and TikTok and ads in ChatGPT, which are maybe known problems that we can design carefully.”

Meanwhile, Altman, former Instacart CEO Fidji Simo (who OpenAI hired as its CEO of applications in early 2025), and seemingly every member of the company’s C-suite have expressed an almost insatiable demand for more compute in interviews.

It’s proven a costly endeavor. OpenAI now has roughly $1.4 trillion in spending commitments on data centers and related infrastructure, drawing questions about how it plans to pay the bills without the benefit of the advertising businesses of its Big Tech competitors, like Google and Meta.

OpenAI also completed its restructuring into a more traditional for-profit, a move Altman said was designed to make it easier to attract future investments.

How the ads will work

As part of the announcement, OpenAI said that free and Go users of the popular AI chatbot would start seeing ads being tested “in the coming weeks.”

Sharing details on the planned test, OpenAI said that ChatGPT’s results “will not be influenced by ads,” the ads will be clearly labeled, and chatbot conversations would remain private and not shared with advertisers.

Paid users of OpenAI’s Plus, Pro, Business, and Enterprise plans won’t see the ads, the company said.

Simo, OpenAI’s CEO of applications, who has previously spoken about her desire to get the ads balance correctly, wrote on X that the most important factor was “ads will not influence the answers ChatGPT gives you.”

While Instacart launched ads during Simo’s time leading the company, she has said OpenAI’s approach would look different.

In November, Simo told Wired that the company would be “extremely respectful” of the deep well of user data, OpenAI has when it came to launching ads.

“If we ever were to do anything, it would have to be a very different model than what has been done before,” she said. “What I’ve learned from building ad platforms is that the thing people don’t like about ads very often is not the ads themselves, it’s the use of the data behind the ads.”

• TikTok has rolled out a new micro drama app in the US and Brazil called PineDrama.
• The app’s feed looks a lot like TikTok’s, but just contains videos from mini drama series.
• The company has been experimenting in the category via a “Minis” section in its main app.

TikTok is making a fresh bet in one of the most promising new media categories: micro dramas.

The company has rolled out a new stand-alone app in the US and Brazil, called PineDrama, dedicated to micro drama series. Micro dramas — also called “verticals” and “mini dramas” — are serialized shows that users watch in roughly one-minute clips.

The most popular micro drama series often include romance or the supernatural, or both — think sordid love affairs with vampires or werewolves. PineDrama’s shows are no exception, with one series, “Love at First Bite,” garnering 18 million views. Its current top three trending shows each had over 100 million views, according to tallies displayed in the app.

On prominent micro drama platforms like DramaBox and ReelShort, viewers can usually watch some episodes of a show for free before being asked to pay $20 a week or more to watch the rest.

PineDrama’s business model is unclear. Right now, users can log into the app with their existing TikTok accounts and watch all episodes of micro dramas like “The Officer Fell For Me,” “The Return of Divorced Heiress,” and “My Unwanted Billionaire Ex” for free. There are also no ads.

“What’s interesting about this one is it’s the first time they’re launching it advertising-free, and it doesn’t yet have a paywall,” said Hernan Lopez, founder of streaming consulting firm Owl & Co. He said PineDrama looks like an experiment and expects it to add monetization over time in the form of subscriptions.

“It’s a sign that vertical video is evolving,” he said. “You’re going to see more announcements this year from many other companies.”

Owl & Co. estimated that the micro drama format had generated $1.3 billion in the US in 2025, mostly from direct payments from viewers. Some investors have been skeptical of the business model, however, because of the high marketing costs required to attract new users.

PineDrama isn’t TikTok’s first foray into micro dramas.

The company earlier released a section in its main app, TikTok Minis, where viewers can binge on bite-sized series without leaving the platform. Many of the same micro drama suppliers on TikTok Minis are also featured in PineDrama, including Dreame, Stardust TV, and ShortMax.

TikTok’s parent, ByteDance, is already entrenched in micro dramas through Douyin, TikTok’s Chinese sister app, and dedicated micro drama apps in Asia, such as Melolo and Red Fruit.

• The Trump administration is leaning on a major power grid operator to bring down prices.
• That includes calling for a new power auction.
• It comes as Trump has pushed for tech companies to “pay their own way” on data centers.

The AI data center boom is driving up electricity prices. President Donald Trump is leaning on a power company to fix it.

Trump and a group of governors from northeastern states are calling on PJM Interconnection, the nation’s largest power grid operator, to hold a new power auction.

The goal is to allow major tech companies to foot the bill for new power plants. According to a release from the Department of Interior, a “coalition of leading technology companies” has committed to funding new generation capacity, though no companies are named.

“Make no mistake, if PJM, this faceless bureaucratic organization that is driving prices up on the American people, does not change, and does not reflect what we are putting forth here today, Pennsylvania will be forced to act, and forced to go at it alone,” Gov. Josh Shapiro said at an event for the announcement, according to Politico.

The Trump administration and the governors are also urging PJM to make a series of other changes, according to a Department of Energy fact sheet. Those changes include:

• Providing 15-year revenue certainty for new power plants;
• Capping the amount existing power plants can charge customers;
• Require data centers to pay for new power plants built on their behalf.

PJM operates in 13 states in the mid-Atlantic and parts of the Midwest. Representatives for the company were not present at the Friday event.

“We don’t have a lot to say on this,” PJM spokesman Jeffrey Shields told Bloomberg on Thursday. “We were not invited to the event they are apparently having tomorrow and we will not be there.”

In a statement on Friday, Shields did not immediately commit to the White House’s directives.

“PJM is reviewing the principles set forth by the White House and governors. The PJM Board’s decision, resulting from a multi-month stakeholder process on integrating large load additions, will be released later today,” Shields said. “The Board has been deliberating on this issue since the end of that stakeholder process. We will work with our stakeholders to assess how the White House directive aligns with the Board’s decision.”

Ari Peskoe, the director of the Electricity Law Initiative at the Harvard Law School Environmental and Energy Law Program, told Business Insider that it’s unclear whether the administration’s approach will work.

“The intent here is correct, which is to appreciate that data centers are driving up prices in PJM, and an attempt to remedy that for consumers,” Peskoe said. “It is meant to isolate data centers from the rest of the market, and, in theory, that seems like a workable approach; but these markets are really complicated, and it’s hard to predict how this all shakes out.”

Trump has been making a broader push for tech companies to pay for the electricity consumed by new data centers in a bid to lower energy costs for consumers.

“My Administration is working with major American Technology Companies to secure their commitment to the American People, and we will have much to announce in the coming weeks,” Trump wrote in a Truth Social post on Monday, adding that tech companies must “pay their own way.”

Microsoft on Tuesday released a plan that includes asking utility companies to set higher rates for the company in order to keep costs down for consumers. Other major tech companies could follow a similar path.

• A European arms maker flew what the Royal Navy called “the UK’s first truly autonomous full-size helicopter.”
• Leonardo’s Proteus demonstrator has completed its maiden flight.
• The move comes after Lockheed Martin subsidiary Sikorsky unveiled an uncrewed Black Hawk.

A new uncrewed helicopter design from European defense and aerospace company Leonardo has taken flight. The move comes after rival US defense firm Sikorsky revealed a pilotless Black Hawk concept last year.

The British Royal Navy on Friday said “the UK’s first truly autonomous full-size helicopter,” the Proteus, completed its maiden flight.

It said that “British aviation history has been made” and that it is “one of the world’s first full-sized autonomous helicopters.”

The Proteus, rather than having a crew in the cockpit or cabin, has sensors and computer systems “driven by cutting-edge software.” Those allow the helicopter to “understand and process its environment, make decisions, and act accordingly,” the Royal Navy said.

A Royal Navy spokesperson told Business Insider that the new design is “a one-off prototype/demonstrator,” explaining that “the Navy and Leonardo will use it for trials and experimentation, but it’s not a production-line machine for everyday usage.”

The navy said in its announcement that Proteus “will serve as a testbed for future hybrid air wings.” It described the helicopter as a step toward having uncrewed aircraft operate alongside crewed platforms.

The Royal Navy operates drones, such as octocopters made by Malloy Aeronautics that can carry supplies between ships, and the Peregrine, which is a “scaled-down helicopter” that does surveillance missions. But, it said, Proteus “eclipses them in terms of size, complexity and above all autonomy.”

Lockheed Martin subsidiary Sikorsky, a key competitor in the helicopter space, unveiled the autonomous U-Hawk, a new uncrewed Black Hawk helicopter, last October.

Pentagon commitments to more autonomous aerial weaponry have pushed US defense companies, both the major players and startups, into intensifying competition to build these systems.

This desire for change is clear in the helicopter space: US Army leadership told Business Insider in July that it wants to invert the current ratio, going to 90% drone and 10% crewed.

Efforts to develop this technology extend beyond the US though, as the Proteus flight shows.

The UK’s autonomous future

The helicopter was designed and built by Leonardo to “unlock the potential of uncrewed aerial systems,” the Royal Navy said.

The aim, it said, is to operate these types of systems alongside crewed aircraft in a “hybrid air wing” combining crewed aircraft, be it traditional rotary wing aircraft or fighter jets, and cheaper autonomous systems like drones.

Leonardo said that the Proteus helicopter concept was designed with a large modular payload bay that can carry two standard NATO pallet loads and that it can perform a host of missions ranging from anti-submarine warfare to airborne early warning. The company said that “with this multi-role design, a single aircraft type can address multiple mission objectives cost-effectively.”

Aircraft like this could play a role in the UK’s future anti-submarine warfare plans under Atlantic Bastion, a new strategy aimed at countering Russian undersea threats. Britain is pushing toward a hybrid naval force designed to detect and respond to submarines and other technologies targeting critical seabed infrastructure.

During its maiden flight, Proteus made two short flights, operating from the Predannack airfield in Cornwall, in the southwest of England, the Royal Navy said.

It said Leonardo, which developed Proteus under a roughly $80 million program, had done ground tests on the helicopter’s sensors and engines weeks prior.

• Yesim Saydan uses over 17 custom GPTs to run her solo-consultancy business.
• She created her ideal team of employees and mentors, including an AI agent inspired by Steve Jobs.
• Saydan worries about the negative impact AI could have, but has learned to embrace it.

This as-told-to essay is based on a conversation with Yesim Saydan, a branding and communication expert in her early 50s, based in the Netherlands. The following has been edited for length and clarity.

When I’m stuck on a business decision or need to come up with a creative idea or strategy, brainstorming usually starts with my Steve Jobs custom GPT.

My solo-consultancy business helps senior executives and established entrepreneurs grow their authority and influence through social media and brand strategies. But scaling that work on my own was challenging.

Before AI, if I wanted to scale the number of clients I could take on, my main option was hiring freelancers for special projects or tasks. I spent a lot of unnecessary time training the freelancers on my specific framework, and it often felt like they didn’t care as much as I did.

When OpenAI launched custom GPTs, everything changed. I used the feature to create over 17 custom GPTs to build my team. Then I thought of my ideal mentors and created custom GPTs of them.

I had to create more than 4 custom GPTs to get good results

One of my first jobs was at Citibank as a project manager on Wall Street, following my move to the US from Turkey for my MBA. That kicked off my 14-year corporate career, during which I worked in New York, Paris, and the Netherlands.

I started my business about a decade ago because I wanted more flexibility in my work schedule. At the time, social media was just starting to take off, and I saw a clear opportunity.

I’d played with AI tools before, but OpenAI’s custom GPTs changed the game. Initially, I envisioned creating my ideal four-person team of agents. I quickly realized AI produces subpar results when it’s overloaded with too many tasks.

Instead, I created a custom GPT for each important task I wanted the AI to perform. That’s how I ended up with more than 17 custom GPTs making up what feels like my perfect team.

I trained my AI team to allow me to focus on the bigger strategy

I can create a custom GPT in five or 10 minutes, but what actually makes it powerful is the training process. I create standard operating procedure documents for each task and client, serving as training materials for my agents that outline my methodology and frameworks.

I have client-specific AI agents trained on each major client’s tone, goals, and past conversations, so I’m never starting from scratch with a task. The training is ongoing. Every time I make a query or upload a document, the agent improves, just like a real employee would.

When I need to communicate or create content in a client’s tone of voice, the draft I end up with is so tailored that it feels like I spent hours perfecting it, when in fact my AI team handled it.

I’ve trained a market researcher, a sales call analyst, a proposal writer, a video scriptwriter, and even a custom GPT to evaluate LinkedIn profiles using six pillars to determine if the current LinkedIn presence builds authority, attracts their ideal client, and establishes trust, clarity, and uniqueness. These free me up to focus on big-picture strategy.

I taught my custom GPT to think like Steve Jobs and mentor me

After creating my ideal employees, I asked myself which mentors I would love to have alive or dead. Steve Jobs is known for his creativity and innovation, and there are numerous videos already online about him; he’s the perfect mentor to create a custom GPT for.

In the instructions, I started with things like, “you are Steve Jobs, you have decades of experience in X, Y, Z, your most important skill is creativity, innovation, thinking out of the box.”

There are two types of video transcripts I trained it on. I uploaded transcripts from videos where he explains his strategies and what he looks for in products. The second approach was training through examples. I found videos showing how he launched products like the iPhone or iPad, so the AI learns from both his thought process and his execution of those launches.

To get it to the level it is now, I spent roughly 40 hours researching and building training assets, including PDFs and other materials the GPT can use as references. I continue adding more whenever I find relevant material, and I now have custom GPTs for Dan Kennedy and Elon Musk as well.

I have to avoid asking certain questions to get the most appropriate responses

The frustrating part with training AI models is that I can give it a lot of information that’s required to have the superpowers of Steve Jobs, but then the AI could take that and produce a lot of different things.

When I prompt it, I avoid asking questions like “What do you think of this idea?” because the AI usually wants to agree with me and please me. Instead, I ask, “On a scale from one to 10, how good is this idea?”

It’s not going to say the idea is bad, but now it might tell me it’s a five. Then I’ll ask, “OK, what would make it a 10?”

That’s usually when it starts drawing on the experience of Steve Jobs that I’ve trained it with. We can go back and forth until I get the most useful and honest feedback possible.

It depends on the task. For more strategic outputs, I usually go through three to five rounds of refinement.

AI scares me, but there’s no turning back now

When a product like NotebookLM was introduced, I started thinking, “Oh my God, this is going to make the entire human race obsolete.” I find AI products fascinating at first, but they can really scare me.

I truly believe we don’t know what the world will look like even a year from now. Sometimes I literally freeze thinking about the impacts, and if everyone will end up homeless, but I usually try to remind myself I’m not God or a higher power, and I don’t know what will happen. This calms me down.

I also realized that AI, by itself, is powerful, but what makes it truly magical is when we combine our expertise and skills with it. Using custom agents as an extension of our brain, rather than a replacement, is what really produces great output.

There’s no turning back from it.