Ex-Twitter CEO Parag Agrawal Launches AI Startup Parallel for 'Web for Machines'
Former Twitter CEO Parag Agrawal has re-emerged in the technology landscape, launching Parallel Web Systems, an artificial intelligence startup aimed at fundamentally reshaping how AI interacts with the internet. Three years after his abrupt departure from Twitter following Elon Musk’s acquisition in 2022, Agrawal’s new venture positions itself at the forefront of building what he terms the “web for machines.”
Parallel’s core mission revolves around developing the tools and infrastructure necessary for AI systems to engage with the vastness of the open web in real-time. This web-scale infrastructure has been engineered from the ground up, specifically designed to empower AI agents to efficiently fetch, verify, and organize information. Agrawal emphasizes that artificial intelligence will soon become the internet’s primary user, consuming data far more extensively than humans ever have. “AIs will use the web far more than humans ever have,” the company stated in its announcement, underscoring its focus on creating systems tailored for machines to undertake complex tasks effectively.
The startup’s product suite includes sophisticated low-level search tools and deep research APIs, which deploy AI agents capable of completing tasks that would typically consume hours of human effort in mere minutes. Parallel asserts that its deep research APIs significantly outperform leading AI models, including OpenAI’s GPT-5, across various benchmarks designed to test complex web search and research capabilities. For instance, on OpenAI’s BrowseComp benchmark, which assesses accuracy in multi-step reasoning and web navigation, Parallel’s deep research APIs achieved an accuracy rate of 58 percent. This compares favorably to GPT-5’s 41 percent and human participants’ 25 percent accuracy, even with a two-hour time limit. Furthermore, in the DeepResearch Bench, a test evaluating the quality and rigor of AI-generated long-form research reports, Parallel secured an 82 percent win rate against a reference set, surpassing GPT-5’s 66 percent.
Parallel attributes its superior performance to its unique infrastructure, which has been meticulously built for AI agents. Every layer of its technological stack—from data crawling and indexing to information ranking—is engineered to optimize how machines consume data, rather than replicating human browsing patterns. This specialized design allows AI agents to process and synthesize information with unparalleled efficiency and precision.
Already, Parallel’s deep research APIs are being utilized by numerous startups, public enterprises, and other AI companies, collectively executing millions of research tasks daily. These applications span a wide range of uses, from AI sales agents researching leads and coding agents synthesizing context from documentation, to investment tools identifying profitable insights from diverse online sources and financial filings. Even insurance companies are leveraging this technology to automate claims verification through web-sourced data.
Parallel Web Systems marks Agrawal’s first major undertaking since his tenure as Twitter CEO concluded in 2022. The startup has successfully raised approximately $30 million in initial funding, drawing support from prominent Silicon Valley venture capitalists, including Khosla Ventures, Index Ventures, and First Round Capital. The founding team comprises experienced engineers who previously contributed to building the foundational infrastructure for the “human web” at companies such as Twitter, Google, Stripe, Airbnb, Chime, Waymo, and Kitty Hawk. Looking ahead, Parallel plans to develop advanced AI agents capable of undertaking the work of entire teams in hours, along with continuous web monitoring systems, event-driven architectures, and SQL-style programmable queries over the internet, further cementing its vision for a machine-driven web.