Month: August 2025

• OpenAI and Anthropic are offering federal employees access to their AI models for little cost.
• The Trump administration also rolled out a government-wide chatbot platform.
• The jockeying shows how the AI race is unfolding in Washington.

While AI giants are breaking the bank to fund the future of the technology, two companies are hoping a single dollar fee can help them gain a foothold in Washington.

OpenAI and Anthropic are charging the Trump administration just $1 per agency to access their leading AI models for the next year.

In another sign of how competitive things remain, OpenAI announced its agreement for ChatGPT Enterprise access on August 6. Less than a week later, Anthropic announced a similar deal for access to Claude for Enterprise and Claude for Government.

Government contracts could be quite lucrative for AI companies. Anthropic already has a deal with the Pentagon that could be worth as much as $200 million.

“Some of these companies are going public, and if they can say their products are being used by government agencies, that boosts their long-term potential,” Darrell M. West, a senior fellow in the Center for Technology Innovation, told Business Insider.

West said companies that aren’t striking these types of agreements risk getting “squeezed out.”

“There are a lot of AI companies now, but that is probably going to narrow in the future — some companies will do well and many are not going to make it — so if you are getting government employees to use your products, it increases the odds that you will be one of the survivors,” he said.

Google might have a similar deal in the offing. Earlier this month, US General Services Administration added the tech giant’s Gemini model to a list of approved AI vendors. After that announcement, OpenAI and Anthropic, which were also added to the list, announced their $1 per agency agreements.

Elon Musk‘s xAi was originally going to strike a similar partnership, but those talks fizzled out after Grok began posting antisemitic content, Wired reported.

Federal employees will have other avenues to use popular AI chatbots.

On Thursday, the US General Services Administration unveiled USAi, a secure platform where federal employees can experiment with AI models at no cost to them. The platform will initially feature models from Google, Meta, Microsoft, Amazon, Anthropic, and OpenAI.

A GSA spokesperson told BI that Anthropic and OpenAI’s nominal fee agreements “are not the same as the USAi access mechanism.”

The partnerships follow President Donald Trump’s unveiling of his AI action plan, a series of policies designed to keep the US at the forefront of the global AI race.

AI companies take maintaining the US’s position seriously. Last year, Anduril and Palantir announced their own effort to outfit the government with AI. OpenAI CEO Sam Altman joined Trump in January to announce Stargate, a $500 billion project that would help rebuild the AI giant’s moat against Chinese competitors.

Anthropic’s agreement also applies to all three branches, underlining that the deal extends to congressional employees.

Despite repeated efforts, Congress has been unable to pass a sweeping AI law. Most recently, lawmakers dealt some in the industry a setback after they stripped out a provision that would have imposed a decadelong moratorium on state-level AI laws from President Donald Trump’s “Big Beautiful Bill.” Its initial inclusion sparked bipartisan opposition. OpenAI, Meta, and Alphabet have all opposed previous state-level efforts to regulate the industry.

As Business Insider previously reported, some lawmakers remain skeptical of using AI chatbots themselves.

• High-school senior Joshua Karoly aims for a tech career despite AI’s rise.
• Karoly began coding in second grade and has developed skills through self-learning.
• He believes AI’s limitations will create opportunities for those with strong coding skills.

Joshua Karoly is a 17-year-old high school senior who lives near Sacramento, California, and wants to pursue a career as a software developer. He hopes that as more people rely on artificial intelligence, he can use his coding skills to land a job despite the technology taking on more work inside companies. The following has been edited for brevity and clarity.

When I was in second grade, I started programming with Scratch, which is super basic block-based programming. I realized I could make games from this. I was like, “That’s awesome. I love games.” Then, I got a book about Python at the library, and I would type in the code from there and was like, “Whoa, I drew a square and made a button that I can click.” Then I moved on to Khan Academy.

I’ve been working my way up from there in terms of complexity. A lot of this was during distance-learning during COVID. When I was supposed to be paying attention in class, I was programming. That’s where I got a lot of my experience. It was very nerdy behavior.

AI might fix one thing, then break something else

When you’re a kid, people always ask you, “What do you want to be when you grow up?” I was interested in computers. So, since the second grade, I would say, “I want to be a programmer.”

As to AI, I knew neural networks existed for a while because I’ve seen people do cool things. Back before OpenAI was very big, they had a song generator that made songs in the styles of classical composers and stuff. I thought that was the coolest thing ever.

At the same time, because of how AI is trained, it’s made to give you output that looks as accurate as possible, even when it’s wrong. A lot of the time, it will tell you for sure that it works, and it looks like it will, but it doesn’t, or it works somewhat.

So, I spend a lot of spend a lot of time debugging AI code, which has been my experience with it. I have been using it now and then to debug my own code or help get an idea of how to figure something out, but generally, I don’t find this code to be the highest quality.

At the beginning of the summer, I did a game jam, which is where you spend a week making a game. I kept having a problem where the code I wrote wasn’t quite detecting where something was. I fed it to the AI, and it was like, “Oh, your problem is this.”

It gave me back code, and that was not my problem. I kept asking it over and over again, trying to help it out, and it didn’t help me. I eventually had to figure it out myself, hours later.

You can get pretty far with just vibe coding, but usually it gets more complicated. As a project gets bigger and bigger, it gets more convoluted as to what the AI can work on. It might fix one thing and then break something else. That can create problems because when your project gets bigger, AI can only focus on one thing. It’s even hard for humans, at least for me. I’m just a teenager.

I’m hoping other young people will focus too much on AI

I’m still not really worried about some big AI supercomputer taking over everything or taking potential jobs. It might play a role in how those jobs are carried out. Maybe there will be less of them.

AI has given me some second thoughts, but there are already so many workers in programming that I don’t know about job security.

I’m sort of hoping that everybody else my age will focus more on AI than they should. So when the bubble bursts a little bit, or maybe when there are jobs AI can’t do as well, then I’ll be the guy for the job because I know how to deal with it.

I’ve only been on this earth for 17 years, and I’m not so great at predicting the future yet. I’m kind of hoping that as people keep relying on AI, people who don’t rely on AI will also be important.

I can still use AI as a tool. I’ve done it, but I try not to rely on it like a lot of people do. For example, in classes, a lot of people use AI, so they don’t really know how to do the thing without it.

AI still isn’t so good at reasoning. It has a long way to go before it starts replacing larger chunks of work I’d like to do or that most programmers do.

The ultimate dream is to run my own company and be my own boss. I enjoy building things of all sorts, especially with code, because it’s abstract. The future isn’t stagnant. It’s not going to stay the way it is now.

Do you have a story to share about your career? Contact this reporter at tparadis@businessinsider.com.

• Tesla’s first car, the 2008 Roadster, was pivotal in establishing the company’s market presence.
• Andrew Lambrecht drove one and noted how features like the lack of power steering stand out.
• Overall, the Roadster demonstrated that electric cars could be fast, efficient, and stylish.

In 2006, Tesla CEO Elon Musk published a blog post titled “The Secret Tesla Motors Master Plan (just between you and me).” Tesla hadn’t built any cars yet, but the post outlined its planned strategy for success.

“So, in short, the master plan is: 1. Build {a} sports car; 2. Use that money to build an affordable car; 3. Use that money to build an even more affordable car; 4. While doing above, also provide zero-emission electric power generation options,” Musk wrote.

Tesla followed this strategy. Its first product was the 2008 Roadster, a $100,000 two-seater sports car. Tesla then built the more mainstream Model S and Model X, eventually paving the way to the Model 3 and Y.

Tesla is almost unrecognizable from its original form when the Roadster launched in 2008. In 12 years, it went from a cash-strapped startup to the world’s most valuable automaker. The Silicon Valley tech firm is now worth over a trillion dollars.

But we cannot forget where Tesla started, and the vehicle that thrust the company into relevance was the Roadster.

The little Roadster that could

When Tesla started, the goal was straightforward: to make a product cool enough to generate interest among the populace. There were many previous attempts at electric cars, primarily in the 1990s, but none of them became mainstream.

Tesla wanted to prove that efficiency, power, and looks were not mutually exclusive pillars to achieve. The result was designing a fast and zingy electric sports car.

The Roadster could thrust itself to 60 miles per hour in just 3.7 seconds and had a range of 244 miles. It could go further than other EVs at the time, squashing the Nissan Leaf’s 73-mile range and the Mitsubishi i-Miev’s 62-mile estimate. Most importantly, it didn’t look or feel like anything electric on the market.

But the numbers only tell part of the story. On the road, it was completely different from anything I’ve ever driven.

Back to Tesla’s roots

When I was in California last summer, my friend Wade Higgins asked me if I would like to drive his Tesla Roadster. Naturally, I cleared out my calendar.

Wade worked at Tesla from 2013 to 2015. In 2014, he sold some of his Tesla stock and picked up a used Roadster he found online. Wade estimates his Roadster could be worth three times what he originally paid, which was $55K.

Still, with Elon Musk’s polarizing right-wing presence and the small number of used Roadsters on the market, current values remain uncertain.

The design

As I reached Wade’s house, the doors from his garage lifted, and the silver 2008 Roadster was unsheathed. I had only seen one on the road prior, so merely being in its presence was a surreal experience.

The first thing I noticed was that the Roadster was tiny — its hood barely made it up to my knees. The Roadster was so small because it was based on the Lotus Elise platform. While many newer EVs are built from the ground up on electric platforms, Tesla didn’t have the resources back in the old days to do so. Therefore, Tesla partnered with the then-struggling Lotus Cars to procure the vehicle’s chassis and help with engineering support.

The result is a 155.1-inch-long car measuring only 44.4 inches in height. For comparison, it’s about 30.7 inches shorter than a Model 3.

This small form factor is excellent for its electric powertrain

Aerodynamics are key to securing the range. Simply put, a smaller car means a smaller frontal area. Less area means the vehicle won’t have to push as much air out of the way when driving.

The car also uses a relatively small battery. Its 53-kilowatt-hour battery pack allows it to achieve its 244-mile range. A smaller battery means less weight and fewer battery cells to keep cool.

Wade’s Roadster is also still on its original battery pack, which, seventeen years later, is still operating normally.

The 2008 Roadster weighs about 2,900 pounds. The low weight means the car doesn’t need mountains of horsepower to go fast. The original Roadster “only” made 248 horsepower, which is far lower than most modern sporty EVs.

The feel

The tight dimensions become immediately apparent when you get behind the driver’s seat. Everything feels miniature. You’ll literally sit shoulder-to-shoulder with the passenger.

Starting the Roadster is also different from any other modern Tesla. There’s no app or sleek keycard — instead, owners have a physical key. Slide it into the receptacle to the right of the wheel, twist it, and you’ll hear a muted electronic hum, an analog ritual for an electric car.

The only attribute the Roadster shares with modern Teslas is the minimalist “TESLA” typeface across the rear. Beneath its sleek exterior, a brutal and raw sports car resides.

Another interesting quirk? There’s no power steering.

At low speeds, the driver receives a complimentary bicep workout. Especially when navigating through parking lots, you must supply a decent amount of force to make the car turn.

The lack of power steering also means there’s no intermediary between the road and the steering column. Every bump, crack, and imperfection in the asphalt’s surface permeates up the steering rack and into your fingertips.

The ride is just as unforgiving. The taught suspension transfers shocks directly into the cabin, creating a bumpy but visceral driving experience. Every jolt and vibration reminds you that comfort was never a priority in the Roadster.

All of these attributes deliver a raw and engaging ride. It was truly one of the most fun EVs I’ve driven in recent years, but I could assume the harshness would get old if I drove this car frequently.

Charging

The Roadster can’t use Superchargers because it launched four years before the first one was built. Therefore, it cannot DC fast charge, meaning you’re limited to slower AC charging.

At home, the Roadster will charge at 70 amps, which is high even for today’s standards. Most EVs top out at 40 or 48 amps. This means the Tesla Roadster can charge at 15.4 kilowatts, adding about 60 miles of range per hour. A zero-to-full charge takes three and a half hours.

Public charging the Roadster adds some complexity. It uses an obsolete charging connector, so a regular Tesla NACS connector won’t work. You’ll need an adapter if you want to charge on a public level 2 station.

The Tesla Roadster wasn’t made to be a mass-market vehicle

The Roadster was made to show the world what electric cars could be.

Despite being 17 years old, the Tesla Roadster still breaks many preconceptions. It’s lightweight, blisteringly quick, and still running strong, thanks to the long-lasting battery pack.

If it hadn’t been for the Roadster, Tesla would likely not be where it is today. I think it’s the most important car of the 2000s.

• Teen prodigy Kairan Quazi joins Citadel Securities as a quant developer in New York City.
• Quazi, who graduated college and joined SpaceX at age 14, chose finance over AI and big tech offers.
• He prized the ambitious, meritocratic culture and fast feedback loops offered in quant finance.

AI labs wanted him. Big Tech did too. But in the end, it was quant finance that won over teen prodigy Kairan Quazi.

The 16-year-old, who garnered worldwide attention in 2023 for graduating from college and landing a job as a SpaceX engineer at age 14, starts this week as a quant developer at systematic trading giant Citadel Securities in New York City.

“After two years at SpaceX, I felt ready to take on new challenges and expand my skill set into a different high-performance environment,” Quazi told Business Insider in an interview. “Citadel Securities offered a similarly ambitious culture, but also a completely new domain, which is very exciting for me.”

In a time when quant trading firms face heightened competition for talent from AI labs like OpenAI, Anthropic, and xAI, landing the engineering wunderkind is a noteworthy win for Citadel Securities, one of the premier destinations in finance.

The sister company of billionaire Ken Griffin’s hedge fund Citadel, Citadel Securities is one of the world’s top high-speed trading firms, handling hundreds of billions worth of stocks, options, currencies, and other assets each day. Using state-of-the-art technology and algorithms, the global market-maker executes trades on behalf of other institutions and retail traders — it handles roughly 35% of retail US stock trades — as well as deploying its own proprietary trading strategies. It generated nearly $10 billion in revenue in 2024 and a record $3.4 billion in the first quarter of 2025, according to media reports.

In his new role, Quazi will work on the firm’s global trading infrastructure, sitting at the “intersection of engineering and quantitative problem-solving” and working with both traders and engineers, a company spokesperson said.

“It felt like a very natural transition from the high-performance culture at SpaceX,” Quazi said.

Why quant finance beat out Silicon Valley

At just 16, Quazi has been amassing accolades and records for most of his life.

Raised in the Bay Area, he jumped from third grade to college at age 9. Quazi interned for Intel Labs at age 10, working on predictive speech generation platforms, and in 2022 interned in machine learning at cyber-intelligence firm Blackbird.AI.

Then, after becoming the youngest ever graduate of Santa Clara University, Quazi took his software engineering degree to the Starlink division at SpaceX.

He wasn’t fetching coffee or working on immaterial projects. Quazi worked on production-critical systems, he says, designing and implementing software that determines where Starlink satellites target their beams, helping ensure fast, reliable internet connections to the company’s millions of customers.

“I had a very broad scope and a lot of responsibility, especially for a junior engineer,” Quazi said.

He added that he was expected to fully own projects, including “developing ideas and presenting them to other engineers and leadership, ultimately implementing them, and monitoring the rollouts.”

With his background and résumé, a long and successful career in Silicon Valley likely seemed a foregone conclusion. And at a time when many of the world’s brightest minds are flocking to AI startups — and courting offers worth hundreds of millions — that would’ve been an understandable next step.

But Quazi declined offers from top AI Labs and tech firms (he declined to specify which companies), as well as other quant firms, to join Citadel Securities.

His rationale for jumping into quantitative trading mirrors the sales pitch the industry and its recruiters often tout: For highly ambitious researchers and mathematicians, it’s a dream scenario with almost instant feedback and responses to your ideas.

“Quant finance offers a pretty rare combination: the complexity and intellectual challenge that AI research also provides, but with a much faster pace,” Quazi said. “At Citadel Securities, I’ll be able to see measurable impact in days, not months or years like many research environments.”

Citadel Securities, in particular, is known for its technological prowess and meritocratic culture, a key selling point for someone who was rejected from jobs — and even booted off LinkedIn — not because of his abilities but rather his age.

He felt Citadel Securities valued his “unconventional” background. “They moved very quickly and didn’t use my age or years of experience to gate-keep opportunities,” Quazi said, adding that the opportunity to engage directly with senior leadership impressed him.

It also helped that Quazi was exposed to finance from an early age, as his mother was an investment banker working in mergers and acquisitions. Moreover, for math and engineering students, quant trading firms loom large on campus as some of the most sought-after post-graduate jobs.

“It’s one of the most prestigious industries you could go into as a computer scientist or mathematician,” Quazi said.

On his own in NYC

Living on your own in New York City as a teenager is nightmare fuel for some kids and their parents.

Quazi isn’t too concerned.

The Bangladeshi-American, who in his spare time likes to read, play video games, rock climb, and hang with friends, among other hobbies, grew up visiting family in New York regularly. He said he values the rich culture, diversity, and energy of the city.

“New York has a very special place in my heart,” Quazi said. “My mom grew up in Astoria, Queens, and I have a lot of family there.”

Another perk: Quazi, who still doesn’t have a driver’s license, will no longer need his mother to drive him to work, as he did for his gig at SpaceX in Redmond, Washington. At least at first, he’ll be living about a 10-minute walk away from Citadel Securities’ midtown office.

After that, he’ll take the subway.