Photo credit: www.theverge.com
Aravind Srinivas is in a competitive struggle with Google to have his Perplexity AI assistant included as a pre-installed app on Android devices. Concurrently, he is redirecting his startup’s focus towards what he foresees as the forthcoming arena in the AI competition: web browsers.
Perplexity is set to unveil its own web browser, Comet, next month. According to Srinivas, “The browser might be the most effective way to develop agents. A browser operates like a containerized operating system, allowing access to various third-party services through hidden tabs and enabling operations on a user’s behalf.”
Other tech companies are also leaning towards this model. OpenAI’s Operator and Google’s Mariner are both designed to utilize a browser for command execution and website interactions. OpenAI is currently believed to be working on its own browser, while Google faces pressure from U.S. regulatory authorities, which have ruled that the company holds a monopoly in the search industry.
Srinivas noted that Perplexity would be interested in running Chrome if it were offered as a separate entity from Google. OpenAI also appears to be exploring similar options.
Despite uncertainties surrounding the future of Chrome, the current antitrust investigations targeting Google create openings for Perplexity to form distribution agreements with Android manufacturers. Recently, Motorola confirmed that Perplexity would be pre-installed on its latest Razr smartphones, potentially broadening access to millions of new users. Srinivas acknowledges that while the integration might not be as deep as he or Motorola envisioned, it’s still a significant step forward for a smaller startup like Perplexity.
“Without the DOJ trial proceedings against Google, this partnership would likely not have materialized,” he explained. “Google has often pressured OEMs, and I’ve encountered situations where telecom companies hesitated to engage with us out of fear that their revenue could diminish if Google took notice.”
A year ago, when speaking with Srinivas, Perplexity had around 1 million users and had secured less than $100 million in funding. Today, however, it boasts nearly 30 million monthly active users and has raised hundreds of millions. Currently, the platform handles approximately 600 million queries monthly, accounting for about 14% of Google’s total query volume.
This article captures insights from a conversation with Srinivas that occurred just before the Motorola partnership announcement, during which they discussed potential collaborations to expand Perplexity’s reach, the strategic importance of browser ownership, and the capabilities of their newly launched iOS assistant among other topics.
This conversation has been condensed for brevity and clarity:
Can you explain how the partnership with Motorola originated and the hurdles posed by Google?
Dialogue with Motorola intensified when they were shown a demo of the Perplexity Android assistant, which was launched in January. They found it to be highly effective—more reliable than Gemini. This led them to consider pre-installing the app and notifying users to make Perplexity the preferred assistant. However, Google intervened, asserting that the phone could not use the Play Store with the official Android version unless Gemini was set as the default.
A partnership like this would not have been possible without the DOJ case against Google. The company’s influence over OEMs is substantial. There have been instances where telecom companies refused to meet with us for fear of repercussions from Google that could affect their revenue share.
Changing the default app requires several steps, indicating Google’s continued stronghold within the Android ecosystem.
Samsung has invested in your company; do you foresee a partnership similar to Motorola’s?
I hope so. We’re still exploring possibilities with them, but we haven’t finalized details regarding defaults or onboarding processes yet.
Your focus on distribution and partnerships for Perplexity’s growth is clear. How do you plan to expand?
We’re open to collaborations with various stakeholders. We’ve engaged with telecom companies and are now looking to extend our reach to OEMs. The browser is next on our agenda, with plans to create versions compatible with both Mac and Windows, and we’ll also aim to engage OEMs in those markets.
Google maintains extensive relationships with OEMs on Android, while Microsoft has similarly restrictive agreements with laptop manufacturers. This presents a challenge we must navigate creatively.
With your new assistant on iOS, users have expressed surprise at its capabilities. Did Apple give you special permissions to manage other apps?
Apple did not grant us such permissions. Our system cannot set alarms, activate low power mode, or control settings like brightness and volume. We utilized the Apple EventKit SDK, which enables access to Reminders, Podcasts, Apple Music, and Apple Maps, among others.
While many criticize Siri for its functionality, it can successfully handle basic tasks like setting alarms. Our focus has been on improving the user experience in areas where Siri falls short, like finding songs or managing ride-hailing services.
Why venture into building your own browser, and what is the timeline?
We’re developing a browser because we believe it offers the best potential for building agents. On mobile platforms, our ability to interact with OS-level functions is limited. A browser serves as a container where we can operate and compare various services side-by-side, such as Uber and Lyft ride prices, or delivery wait times.
We aim to create a comprehensive OS-level agent. Browsers provide a flexible framework for users to interact with numerous online services seamlessly. The goal is to transition from merely answering questions to enabling actionable steps, which is why we see the browser as a crucial next step.
Publishers have shown dissatisfaction over your content scraping practices. Have you made progress in strengthening those relationships?
While there’s still work to be done, our relationship with publishers has significantly improved since last time we connected. We respect robots.txt protocols for scraping and have engaged third-party data providers instead of scraping data where it’s not permissible.
Your reported funding round could set your valuation at $18 billion. What are your plans for utilizing these funds?
To build reliable agents, we must utilize cutting-edge reasoning models now rather than waiting for costs to decline. Our strategy involves rolling out our technology to a vast user base quickly, allowing us to gather data to refine our models and reduce expenses.
What is the current status of your bid for TikTok? Have you had any recent discussions with the White House?
I remain interested but recognize my chances are not optimal. It’s not a funding issue; numerous backers were willing to support me. The concerns we encountered from ByteDance were more about their desire to retain control over their algorithms and ownership of the app.
Do you see challenges with competing against ChatGPT, given its growing user base and personalized capabilities?
I believe their approach includes offering a “Login with ChatGPT” option across external applications to aggregate user data. However, convincing third-party apps to adopt this could be challenging.
Our strategy prioritizes seamless user access, allowing people to remain logged into their preferred services. By building a browser, we can reach applications on behalf of users more effectively.
Success in personalized assistance will depend on context—the ability to understand users’ transactions, commerce history, and browsing data collectively. This data integration is the real challenge before us.
The browser seems to be your ultimate goal in building out this platform.
Indeed, it’s just the beginning; the real prize lies in developing full operating systems for Windows, Mac, Android, or iOS. The browser serves as a starting point, but the OS remains the ultimate battlefield for innovation.
OpenAI is facing an urgent demand for more computing resources: At a recent closed-door meeting with investors in San Francisco, OpenAI’s leadership raised concerns regarding their access to the computing power necessary to sustain ChatGPT’s rapid evolution. This demand appears more urgent than the pursuit of profitability at this time, indicating that, despite advancements in operational efficiency by companies like DeepSeek, the costs associated with state-of-the-art models continue to rise. (The Information provides an insightful breakdown of OpenAI’s latest financial projections, forecasting that revenue from non-paying users and alternative products may reach $25 billion within four years.)In brief: Elon Musk announced his intention to step back from DOGE; Apple and Meta faced initial fines under the EU’s Digital Markets Act, which the U.S. labeled “economic extortion”; Meta implemented layoffs in its Reality Labs division; Intel announced plans to streamline its teams while adopting a four-day workweek; and Google requires certain remote teams to return to the office part-time.
Important career updates and job opportunities:
Discord co-founder and CEO Jason Citron is stepping down to dedicate time to playing Final Fantasy VII Rebirth, while Humam Sakhnini, formerly of Activision Blizzard, will assume the CEO role at Discord as the company prepares for its IPO. The future direction under Sakhnini remains to be seen: whether he will cultivate its gaming roots or expand its mission further.
Sam Altman has stepped down as chairman of nuclear-energy firm Oklo, a move he suggests is aimed at avoiding potential conflicts as the company pursues AI partnerships. While Oklo is publicly traded and subject to enhanced governance, further scrutiny may arise regarding Altman’s involvement with other startups.
Ranjit Desai and several other leaders have transitioned from Apple’s Vision Products Group to bolster Siri’s development. DeepSeek is on the lookout for product leaders in China to advance a “next-generation intelligent product experience.” Google’s Gemini app now boasts 35 million daily and 350 million monthly users, while Microsoft Copilot sees flat growth with 20 million weekly users.
Dario Amodei wrote about the “urgency of interpretability” in the current AI race, and Kevin Systrom has emerged from retirement to comment critically on Meta. The competition for talent continues amidst near-IPO startups like Figma, while X (formerly Twitter) is experiencing sluggish ad growth yet is leveraging data sales to Grok. Amazon is also launching “Project Greenland” to deal with GPU shortages.
Phoebe Gates, the daughter of Bill and Melinda Gates, has launched an AI shopping application, and 60 Minutes recently featured an interview with Google DeepMind CEO Demis Hassabis. Lastly, the intrigue surrounding Infinite Reality, a metaverse startup with a claimed valuation of $15 billion, continues to grow.
Source
www.theverge.com