Parag Agrawal, the former CEO of Twitter (now X), is making a return to the tech world with his new AI startup, Parallel Web Systems. The company is focused on building infrastructure that allows AI systems to interact with the internet in real-time. Parallel's goal is to create tools and APIs that enable AI agents to efficiently retrieve, verify, and organize web content.
Agrawal's departure from Twitter in 2022, following Elon Musk's acquisition of the company, was abrupt. After leaving Twitter, Agrawal spent time reading research papers and sketching out ideas in Palo Alto coffee shops before founding Parallel Web Systems in 2023. Agrawal felt many of the other roles being offered to him involved simply "cleaning up shit" and was clear his next step would be in artificial intelligence.
Parallel Web Systems aims to build a "web for machines". The company's mission is to create web infrastructure specifically designed for AI, enabling machines to consume, process, and act on information more effectively than humans. Agrawal believes that AI agents will become the primary users of the web, conducting research, analysis, and automation at scales far exceeding human capabilities. According to Parallel, AI agents will be the primary customers of the web moving forward, using the web far more than humans ever have.
Parallel's suite of products includes low-level search tools and deep research APIs that deploy AI agents to complete hours of human work on the web in minutes. The company claims its deep research APIs can outperform leading AI models, including OpenAI's GPT-5, on complex web search and research tasks. For example, Parallel's deep research APIs achieved 58% accuracy on OpenAI's BrowseComp benchmark for accuracy in multi-hop reasoning and web navigation, while GPT-5 recorded 41% accuracy. Human users scored 25%.
The company's infrastructure is specifically designed for how computers consume information, unlike conventional search algorithms intended for human browsing. Parallel's platform allows AI applications to search the public web in real time, verify the accuracy of the information, and package the results into clear responses. The system includes eight "research engines" designed for different needs. The fastest engine can deliver results in under a minute, while the most advanced can spend half an hour digging for highly detailed answers.
Parallel has already raised $30 million from investors, including Khosla Ventures, Index Ventures, and First Round Capital. The company's founding team includes engineers with experience from Twitter, Google, Stripe, Airbnb, and Waymo. The team will look to develop long-horizon AI agents that can complete the work of entire teams in hours, continuous monitoring systems that track the web for signals, event-driven architectures, and SQL-style programmable queries over the web.
Agrawal envisions a future where individuals deploy dozens of AI agents to perform online tasks, a shift he believes could occur as early as next year. He hopes that Parallel, not Twitter, will define his career as he steps back into the Silicon Valley spotlight.
Parallel is positioned as a cornerstone of the upcoming web architecture. The company intends to diversify into programmable web inquiries, continuous monitoring systems, and long-horizon AI agents in the near future. Early applications include financial tools that analyze regulatory filings, coding assistants that synthesize paperwork, and AI sales agents that conduct lead research. Parallel is already powering millions of research tasks daily.
