TLDR: Former Twitter CEO Parag Agrawal’s AI startup, Parallel Web Systems Inc., has launched a cloud-based research platform that reportedly outperforms OpenAI’s GPT-5 and even human capabilities in deep web research benchmarks. The company, backed by $30 million in funding, aims to build the foundational infrastructure for AI agents to interact with and utilize the public web at an unprecedented scale.
Palo Alto, CA – August 17, 2025 – Parag Agrawal, the former Chief Executive Officer of Twitter, has officially unveiled his new artificial intelligence venture, Parallel Web Systems Inc., with a bold claim: its advanced research platform has surpassed OpenAI’s GPT-5 in deep web research benchmarks. The company, which publicly launched its cloud-based platform on August 14, 2025, is poised to redefine how AI agents interact with and extract information from the vast expanse of the internet.
Parallel’s flagship offering, the Deep Research API, features a top-tier engine known as Ultra8x, which has reportedly outperformed GPT-5 by more than 10% across two challenging benchmarks. The company asserts that its technology also surpasses human performance in these complex research tasks. This breakthrough positions Parallel as a significant player in the burgeoning field of AI infrastructure, focusing on enabling AI models to conduct large-scale, verifiable online research.
Founded in 2023, Parallel Web Systems has rapidly assembled a 25-member team in Palo Alto and secured $30 million in funding from prominent investors, including Khosla Ventures, First Round Capital, and Index Ventures. Agrawal’s vision for Parallel is to construct the ‘search infrastructure for the web’s second user’ – artificial intelligence. He posits that AI will utilize the web ‘a thousandx a millionx more than humans ever have,’ necessitating a fundamental transformation of the internet itself to cater to AI’s unique needs.
The platform offers a suite of eight distinct AI research engines, each tailored for varying computational demands, from rapid queries completed in under a minute to extensive, 30-minute deep explorations. Unlike traditional search engines, Parallel operates its own dedicated crawler and index, providing AI models with direct, real-time access to public web data. A key differentiator is the platform’s ability to provide confidence scores and detailed citations with its prompt responses, enhancing the reliability and verifiability of the retrieved information. Furthermore, it allows for the organization of gathered data into specific, structured formats, such as spreadsheets.
Also Read:
- OpenAI Unveils ChatGPT Agent: A New Era of Automated Personal and Professional Assistance
- The Dawn of Agentic Browsers: Opera and Perplexity See Web Browsers as Central to AI’s Future
Agrawal’s return to the tech forefront follows his unceremonious dismissal from Twitter by Elon Musk in October 2022. Eschewing a break, he immediately immersed himself in building Parallel, driven by a conviction that his next venture had to be in AI, especially after OpenAI’s ChatGPT launched shortly after his departure. He expressed a desire for Parallel, rather than his tumultuous tenure at Twitter, to define his career. The company envisions a future where AI agents can perform complex tasks like assisting programmers by pulling code from GitHub repositories, aiding retailers in competitive analysis, and automating intricate enterprise workflows, all powered by Parallel’s robust web research capabilities.
