Former Twitter CEO Parag Agrawal has launched a new AI startup called Parallel Web Systems, and it’s worth paying attention to. The company is building tools that let AI access the live web, not just rely on old data it was trained on. Most AI models, even ones like GPT-5, struggle when they need up-to-the-minute information. Parallel changes that by letting AI agents search, verify, and organize live web content automatically. It makes their results more accurate and reliable. Tests show it beats GPT-5 by more than 10% on deep research tasks, and it’s already processing millions of queries every day.
Twitter CEO’s New Startup And Seed Funding
Parallel was quietly started years before, after Agrawal left Twitter, and it currently has a team of about 25 people. They just closed a $30 million seed round led by Khosla Ventures, Index Ventures, and First Round Capital. The funding is going into building more powerful “research engines,” expanding the company’s web-crawling systems, and growing its platform for both enterprise companies and developers. This move feels like both a personal comeback for Agrawal and a bet on the next generation of AI tools.
Depth With Research Engines
The technology is designed to balance speed and depth. Parallel has eight specialized research engines. One engine returns results in under a minute for quick questions. Another, called Ultra8x, can take up to 30 minutes on a single task to go deep, digging through multiple sources and connecting information. On industry tests, Ultra8x consistently outperforms GPT-5 and even human researchers.
APIs Enable Real Time AI Applications
Parallel offers three main APIs. The Task API handles complex research for analysts or AI systems. The Search API gives fast results for chatbots or coding tools. The Chatbot API is built for conversational AI that needs answers quickly. In practice, this means AI can pull live code from GitHub, track competitor products for retailers, or compile customer reviews into organized datasets. Early customers already include some of the fastest-growing AI companies.
Vision For Multi Agent Real Time AI
The bigger idea behind Parallel is building the web for AI, not just humans. Current websites are designed for people, which limits how well AI can use them. By creating AI-friendly web systems, Parallel removes this “accuracy ceiling” and lets AI agents work at scale. Developers can choose how much computing power to use depending on the task, and every result comes with sources and confidence scores. That transparency is especially important for regulated industries.
Agrawal’s vision goes beyond single queries. He imagines dozens of AI agents working at the same time for each human user, monitoring changes online, running multi-step projects, and even interacting with live data in a programmable way. With $30 million in seed funding, strong performance against GPT-5, and enterprise-level security, Parallel Web Systems looks positioned to become a core tool for any AI that needs real-time information. For an Ex Twitter CEO, this is a strategic, forward-looking move that could reshape how AI interacts with the web.
About The Author
