This week, OpenAI unveiled an AI-enhanced web browser named ChatGPT Atlas, leading me to ponder: Is it finally time to move on from Safari?

That update was a topic of conversation as Max Zeff, Sean O’Kane, and I explored the browser ecosystem — including some lesser-known options — in the most recent episode of the Equity podcast. Nevertheless, it appears that none of us will be making any significant changes in the near future.

Firstly, Sean pointed out that numerous companies have attempted and ultimately failed to dethrone the dominant browsers, primarily due to their inability to generate revenue solely from the browser. Naturally, this is less likely to hinder OpenAI, given its continuously expanding financial backing.

On the other hand, Max has actually experimented with Atlas and other browsers that claim AI agents will handle tasks for users. He noted that the efficiency improvement is minimal at most. Often, users find themselves watching the agent “navigate a website” — is that what average users genuinely desire? Furthermore, there are considerable security concerns.

View a summary of our dialogue below, refined for brevity and clarity.

Anthony: I still use Safari, but regarding search engines, which are linked to browsers, I’ve been trying to explore non-Google [alternatives] because I’m simply weary of seeing all the genAI content at the top of my search results.

Additionally, there’s the issue of: If these AI browsers gain popularity, what consequences will that have for the concept of the open web overall? You can still access web pages, but it isn’t far-fetched to argue that a website will become increasingly less significant as more of our browsing experiences are managed by these AI platforms and chatbots.

Max: This has certainly been a significant topic of discussion: What will the agentic web entail? It poses an intriguing question. People have made efforts to devise numerous solutions to reach this future they anticipate. 

There’s a certain nostalgic aspect that reminds me of previous technological trends where it’s like, “Okay, but what is the real experience? What tangible benefits does a consumer gain by using one of these tools?”

At the moment, it’s just not particularly enticing. I’ve tested ChatGPT Atlas and Comet, and the most favorable evaluation is that they provide a marginal efficiency increase — making you slightly more productive.

However, most of the time with these tools, you end up watching them navigate a website, performing tasks that I’m unlikely to do in reality. For example, I might have it search for a recipe and add all the ingredients to Instacart. I’ve never done that. It seems like a familiar example tech enthusiasts often use, but I’m not convinced it’s something people frequently do.

There’s a substantial disconnect, with the tech industry currently proclaiming, “We’re developing all these tools for the agentic web,” but what incentive is there for ordinary users to adopt this? I remain uncertain.

Sean: I haven’t used any of those [AI browsers], largely because I’m still quite traditional when it comes to searching and browsing — much of my work involves seeking documents, which naturally requires navigating through various parts of web pages I know well, often employing Boolean searches on Google. Perhaps I’ll give them a shot someday if Google indeed decides to phase out Boolean search, which seems like a possibility, but we’re not there yet. 

What intrigues me about these AI browsers is that we’ve seen other firms attempt to penetrate the browser market and consistently fail, as it is exceedingly challenging to profit from a browser as a standalone product. Some have attempted to charge initially, managing to sustain for a while, but it’s ultimately unsustainable when competing with Safari or Chrome or even Firefox. 

What’s fascinating to me is that we now have companies with virtually unlimited resources, allowing them to endure as long as necessary since they aren’t currently focused on monetizing these products. Eventually, they likely will, but OpenAI can afford not to monetize this for the next few years; they can simply release it and let it evolve.

[embedded content]

A few Thursdays back, I woke up around 4:30 a.m. to a bewildering Instagram DM.  

Rizzbot, a child-sized humanoid robot created by Unitree Robotics and boasting a substantial social media presence — over 1 million followers on TikTok and more than half a million on Instagram — had sent me an image: it was giving me the finger. 

No text. No context. Just a robot with its middle finger extended.  

While I was taken aback, a sinking feeling led me to suspect the reason. A couple of weeks prior, Rizzbot — or the individual behind its Instagram account — and I discussed a potential story. I found the account fascinating: a humanoid trotting around Austin sporting Nike dunks and a cowboy hat. It’s recognized for roasting, but also for flirting and havingfun. The term Rizz is derived from the Gen Z slang word rizz , which means charisma. 

I was fascinated by the account’s increasing popularity. Humanoids usually make people uneasy. Concerns about privacy and job loss abound. Online, individuals hurl insults at them, most notably referring to them as “clankers.” Meanwhile, in the robotics community, experts are debating their optimal applications.  

I viewed Rizzbot as a paradigm, helping others feel at ease in engaging with a humanoid. 

Rizzbot consented to an interview, prompting me to reach out to experts to explore the future of humanoids as I prepared for the story. Two weeks after my first DM with Rizzbot, I informed it I would finally deliver interview questions on the ensuing Monday or Tuesday.  

But life intervened, and I overlooked my own deadline. I was all set to send the questions first thing Thursday a.m., and I thought it was no big deal.  

Too late. In the early hours of Wednesday night, Rizzbot dispatched that image. The message was clear: You broke your promise, so get lost. 

I didn’t give up, though. I apologized to the robot (or its operator?) for the delay and assured I would send the questions right at the start of business hours. But when I attempted a few hours later, I was greeted with “user not found.”

The robot had blocked me.  

Did I activate a fail-safe? 

My friends found it hilarious that I was flipped off and blocked by Rizzbot, considering I had been talking excitedly about this story for weeks.  

“LOL Rizzbot served you,” one friend texted me.  

“YOU ARE FIGHTING WITH A ROBOT LOLOLOL,” another remarked. I attempted to reach out to Rizzbot on TikTok, a move one friend called desperate. But what choice did I have? I had pitched the story to my editor, invested hours in research, and — despite this conflict — Rizzbot would still be an intriguing subject for TechCrunch’s tech-savvy audience. 

While my friends were laughing, I spiraled into a state of despair. Not only was my story effectively dead, but I had also become the girl who got blocked by a dancing robot.  

My colleague Amanda Silberling volunteered to assist me. She reached out to the Rizzbot account to inquire why I was blocked. Rizzbot provided a terse reply: “Rizzbot blocks like he rizzes — smooth, confident, and with no regrets.” It subsequently sent her the same middle finger image it sent me. I thought: Wow, I wasn’t even unique enough for a different gesture of discontent.

Then, one friend presented a chilling perspective I hadn’t even considered. “It wasn’t a human reaction. I’m concerned for you.” It appeared I had inadvertently created my first robotic adversary, and the AI uprising had only just commenced.  

Or had I? Was I truly fighting with a human? 

I discovered that Rizzbot’s true name is actually Jake the Robot.  

Its owner is an unidentifiable YouTuber and biochemist, as per reports. The robot itself is a generic Unitree G1 Model — manufactured in Hangzhou, China — and anyone can acquire one for $16,000 to over $70,000.  

Rizzbot was trained by Kyle Morgenstein, a PhD candidate at the UT Austin robotics lab. He collaborated with a team for about three weeks, teaching the robot to dance and maneuver its limbs. Although much of the robot’s actions are preprogrammed, it is run via remote control, with its actual owner — apparently not Morgenstein — nearby to direct it.  

If I had to speculate about the technology powering the robot — informed by discussions with Malte F. Jung, an associate professor at Cornell University who studies information sciences — it seems someone activates the robot’s actions, captures an image of whoever is engaging with the robot, processes it through ChatGPT or a similar language model, and then utilizes a text-to-speech feature to roast or flirt with the individual.  

“The robot inverts the narrative of people mistreating robots,” Jung stated. “Now the robot gets to mistreat people. The real product is the performance.”  

Morgenstein shared with other outlets that the actual owner of Rizzbot just enjoys entertaining people and showcasing the delight that humanoids can provide. 

It remains unclear who manages the Rizzbot social media accounts, although when Rizzbot sent that image to Silberling, it also dispatched an error message — probably an accident — indicating it was out of GPU memory. This message suggested that an AI agent likely runs that account and may be auto-generating DM responses. It also implied that Rizzbot has only 48GB of memory.  

“What makes you so sure it was ever a human?” my coder friend queried about the manager of the Instagram account. 

In this AI era, anyone capable of training a robot probably possesses the skills to link an LLM to Instagram DMs. My blocking might have even been a fail-safe, my coder friend suggested, meaning I inadvertently triggered it myself by DM’ing in the early hours — even if it was a response.  

However, there are indications that a person is in charge of Rizzbot’s social media: initial typos appeared in its first DM reply to me when I initially sought an interview. 

Even so, unless Rizzbot discloses whether his social media manager is another bot (which seems unlikely in light of our dispute), I will probably never find out. Perhaps it doesn’t matter. 

“If they invested $50,000 in a bot and a couple thousand for a 48GB memory machine, I wouldn’t dismiss anything,” my coder friend pointed out. “They’re clearly committed to the performance.”  

It’s still robot brain fog 

Rizzbot’s TikTok page alone has garnered over 45 million views. One clip features Rizzbot pursuing individuals on the streets, while another shows it colliding with a pole and tumbling into the street. A viral clip, presumably modified by AI, depicts Rizzbot being run over by a vehicle.  

“Honestly, it seems amusing,” one founder friend told me, labeling the viral clips “robot brain fog.” He asserted that the AI is basic, but the robot’s concept is a “humorous fusion” of internet absurdist humor and the lightness that much of social media lacks nowadays. “It interacts with people in a unique manner.” 

My Rizzbot exploration continued to provoke thoughts about the place of humanoids in our culture. Every sci-fi film I’ve ever viewed — from “Blade Runner” to “I, Robot” overwhelmed my memory. How concerned should I be now that I’ve created my first humanoid enemy?  

“Performance truly appears to be the main application for these kinds of robots,” Jung shared, indicating that Rizzbot resembles “a contemporary take on street performance with a hand puppet.”  

“Often, hand puppets have a sarcastic edge,” he added.  

Apart from Rizzbot, he cited the Spring Festival performances in China, where humanoids dance alongside humans, while in San Francisco, spectators visit the boxing arena to watch robots throw punches.  

“Robots will evolve into primary mass entertainment figures, including performers, dancers, singers, comedians, and companions,” Dima Gazda, the founder of the robotics company Esper Bionics, remarked, noting that humans will transition to niche, top-tier talent. “As robots acquire grace and emotional insight, they’ll blend into performances and interactive experiences more seamlessly than humans.”  

Fortunately, at this moment, dancing robots seem challenging to scale en masse, according to Jen Apicella, executive director at the Pittsburgh Robotics Network. Therefore, I don’t have to dread the escalation of this conflict to, for example, an army of dancing, rizzing robots personally arriving at my door. Not that such a thought crossed my mind. 

It has been over a week since I was blocked, and I find myself reflecting on the joy I experienced watching Rizzbot pursue people on the streets. My favorite video featured a woman dancing on Rizzbot. A crowd gathered around the spectacle; people appeared genuinely entertained, perhaps eager for their own chance to dance on a robot.  

I always joked with my friends about wanting to keep robots on my good side in case a revolution occurred. But as I wrote this article, I found myself almost embroiled in another AI conflict — this time with Meta AI, which I had never engaged before. I accidentally initiated a conversation with Meta AI while searching for my old chats with Rizzbot on Instagram.   

Meta’s bot responded, “Yoo, what’s good fam? You referring to me as Rizzbot? 🤣 What’s going on?”   

I decided it was time to log off.

On several weekends, 21-year-old student Lavanya Jain accesses the BlaBlaCar app to secure a ride from Noida, located on the edges of New Delhi, to his residence in Kandhla, a quaint town in the northern part of India’s Uttar Pradesh region. The trip, covering 120 kilometers, costs him roughly ₹500, approximately $6. This is a mere fraction compared to the ₹1,500–₹2,000, or $17–$23, he would incur for a private taxi.

“If you’re seeking a quick, efficient, budget-friendly, and comfortable method to travel—and enjoy conversations—BlaBlaCar is definitely worth checking out,” Jain shared with TechCrunch, noting that he has utilized the app around 40 to 50 times in the last two years.

Jain is among millions of Indians who are opting for long-distance carpooling as a cost-effective and more social means to travel between cities. This trend has made India BlaBlaCar’s largest global market, estimating around 20 million riders this year—nearly a 50% increase from last year. According to this prediction, BlaBlaCar’s India market would exceed the 18 million riders anticipated in Brazil and its native market of France.

The turnaround is remarkable for a company that closed its India office in 2017 due to low demand.

The growth largely occurred without marketing efforts or a localized workforce, instead fueled by word-of-mouth referrals, increasing mobile internet accessibility, and the growing trend of digital payments and car ownership among India’s middle class.

India boasts over 700 million smartphone users and has experienced a significant upsurge in digital payment methods, now constituting over 99% of all transactions nationwide.

Central to this transformation is the government-supported Unified Payments Interface (UPI) system, which facilitated approximately 19.6 billion transactions totaling around ₹24.9 trillion (about $284 billion) just in September. Vehicle sales have risen correspondingly, with 4.73 million cars sold in 2024, increasing from 3.87 million the previous year—an uptick of 5.2% year-on-year and a new record.

Additional contributors to BlaBlaCar’s swift advancement in India encompass the nation’s insufficient public transport relative to its over 1.4 billion populace, along with the ongoing growth of road development enhancing links between minor towns, rural regions, and larger cities.

“We have numerous instances from users stating, ‘Previously, I would fly to a destination or take a train, or sometimes not travel at all—now, I can conveniently drive. It takes three hours, and it’s a delightful journey,’” shared Nicolas Brusson, co-founder and CEO of BlaBlaCar, in an interview.

BlaBlaCar first penetrated the Indian market in early 2015, establishing a local branch in New Delhi. The firm quickly encountered strong competition from Uber and local counterpart Ola, both of which were trialing carpooling options and extensively advertising them. (These companies ultimately suspended their carpool services during the COVID-19 lockdowns.)

Finding it challenging to establish a foothold, BlaBlaCar ceased its local operations in 2017. Nevertheless, the app remained operational—and by 2022, usage began to rise again. Since that point, it has soared from 4.3 million users in 2022 to a forecasted 20 million this year.

This year, BlaBlaCar has averaged approximately 1.1 million active users monthly in India, peaking at around 1.5 million in August. About 75% of these users are passengers, while the other 25% are drivers. Currently, India constitutes approximately 33% of the global carpooling users for BlaBlaCar, the company reported.

In terms of trips, BlaBlaCar achieved its healthiest growth in India, reporting 13.5 million trips completed as of September 30, an increase from 9.1 million during the same timeframe last year. Brazil remained slightly ahead, with 14 million trips this year compared to 11.7 million in 2023, while France held third place with 5.6 million trips, remaining relatively unchanged from the previous year.

“Our focus has shifted away from our initial markets in Western Europe toward regions like Japan, Turkey—and increasingly, India,” Brusson conveyed to TechCrunch.

Although BlaBlaCar does not currently generate income from India, drivers utilizing its platform earned around ₹713 million (approximately $8 million) just in August, as reported by the company. On average, drivers make about ₹390 (around $4) per seat in India, with the average trip distance being 180 kilometers (approximately 112 miles).

In contrast, average earnings for drivers are approximately €15 (around $17) in France and about €6.5 (around $7) in Brazil, despite comparable trip distances in India and Brazil, which are shorter than France’s average of roughly 250 kilometers (around 155 miles). BlaBlaCar noted that this discrepancy is attributed to lower local purchasing power and expectations for cost-sharing in India.

Almost 70% of BlaBlaCar’s users in India are aged between 18 and 34, and about 95% of interactions occur through its mobile application. Approximately half of the rides in India take place along the country’s 15 busiest intercity routes, with the other half coming from outside the top 150 corridors—indicating growing acceptance beyond major metropolitan areas and into smaller towns. Among the busiest routes include Pune–Thane and Pune–Nashik in Maharashtra, Bengaluru–Chittoor in Andhra Pradesh, and several other connections between mid-sized urban centers.

“No hurry” to initiate monetization

In spite of this growth, BlaBlaCar is not planning to introduce monetization in India in the near future.

“We are not rushing to start charging fees or earning revenue in India. Our focus is on increasing usage, and we have the experience from multiple markets to back that up,” Brusson stated to TechCrunch.

Nonetheless, BlaBlaCar intends to establish a local office in India with its first recruit expected by the end of this year or early the next, Brusson indicated.

BlaBlaCar does not see ride-hailing services like Uber and Ola as direct competitors within India. Brusson characterized them as “demand-driven” services, while he pointed out that BlaBlaCar is “supply-led.” The company believes that individuals using their own vehicles—or opting for easily accessible trains and buses—are its primary alternatives.

Obstacles on its path to success

BlaBlaCar continues to encounter several challenges in India.

State regulations regarding carpooling are unclear, which has led to the service being scrutinized in certain cities. Some users, including Jain, have voiced concerns over the difficulty in reaching BlaBlaCar’s customer support, which often relies on automated responses. The company informed TechCrunch that they maintain a “blended model,” with a local outsourced team managing most everyday inquiries and a smaller group at their Paris base overseeing intricate issues and quality assurance.

BlaBlaCar has launched an ID Check feature in India to authenticate users’ identities using government-supplied documents—a tool that has since been introduced globally. Nevertheless, TechCrunch discovered that users can still book or list a ride even if their ID verification is incomplete.

“This choice was intentionally made to facilitate engagement for new members on the platform,” the company responded. “ID verification is just one component of our extensive trust and safety strategy; we do not depend on a single feature, but rather on multiple, interconnected systems that collectively foster confidence within our community.”

The company emphasized that over 70% of rides in India are conducted by drivers who have undergone government ID verification. BlaBlaCar also showcases user reviews and ratings and authenticates accounts via phone numbers and email addresses.

“We strongly encourage members to finalize all verification processes, as profiles that are thoroughly verified—with photo and ID—greatly enhance the likelihood of securing carpool companions. Profiles that lack these elements tend to attract fewer bookings,” the company stated.

Some users of BlaBlaCar in India have also expressed dissatisfaction when drivers or riders cancel trips at the last moment, sometimes even after arriving at the meeting point. Furthermore, the app currently lacks a live location-sharing feature, which Jain pointed out limits its utility for those trying to arrange rides for family members or friends.

BlaBlaCar has modified its service offerings to better cater to Indian users, introducing features like “meeting-point logic” to facilitate coordination. Unlike countries such as France, where fixed carpooling zones are available, India does not have designated pickup spots. Drivers and riders typically agree to meet at convenient locations along the journey—like a petrol station or near a highway exit. The app now recommends and displays these sites using a combination of machine-learning algorithms and user feedback, assisting in minimizing detours and aligning with India’s existing infrastructure, as stated by the company.

Globally, BlaBlaCar anticipates accommodating approximately 150 million passengers this year, which includes users of its bus services operational in markets like France but not yet in India. As BlaBlaCar expands internationally, India’s unexpected growth has positioned it at the core of the company’s next phase of expansion.