Former Twitter CEO Parag Agrawal Launches AI Startup Parallel, Outperforming GPT-5 in Web Research

Former Twitter CEO Parag Agrawal's AI Startup Parallel Beats GPT-5 in Deep Web Research

Nearly three years after his abrupt departure from Twitter (Now X) following Elon Musk's takeover, Parag Agrawal has launched a new AI startup, Parallel Web Systems, where he serves as founder and CEO. Backed by prominent venture capital players, including Khosla Ventures, Index Ventures, and First Round Capital, the startup has raised approximately $30 million in funding. Parallel's founding team includes engineers who previously worked at Twitter, Google, Stripe, Airbnb, Chime, Waymo, and Kitty Hawk. Parallel aims to empower AI systems with the ability to interact with the public web in real time. Instead of relying on static data, the system enables AI agents to fetch, verify, organise, and even assess the confidence of the information they present, essentially giving AI its own browser. "Agents are going to be the primary customers of the web going forward. They will use the web a lot more than humans ever have," said Agrawal. Parallel has also released its deep research APIs that outperform both humans and leading AI models, including GPT-5, on benchmarks testing complex web search and research tasks. The company reported that on OpenAI's BrowseComp benchmark, which measures accuracy in multi-hop reasoning and web navigation, Parallel achieved 58% accuracy compared to 41% for GPT-5 and 25% for human participants working with two-hour time limits. On DeepResearch Bench, which evaluates long-form synthesis across 22 fields, Parallel recorded an 82% win rate against a reference set, compared to 66% for GPT-5. "AIs will use the web far more than humans ever have. As much as we might anthropomorphise AIs, they operate differently from humans on the web," the company said in its announcement. "At Parallel, we are building for the web's second user. We are creating systems and infrastructure for AIs to use the web effectively for completing complex tasks. " Parallel said its APIs are already powering "millions of research tasks daily" for startups, public enterprises, and AI companies. According to the company, coding agents use the platform to search documentation and debug, while enterprises deploy it to automate workflows such as insurance claims processing. Parallel's infrastructure was built specifically for AI agents, with each layer of its stack, from crawl and index to ranking, engineered for machine consumption rather than human browsing. The company said this approach allows "declarative interfaces" where agents specify what they need, without detailing how to retrieve it. Looking ahead, Parallel said it is developing capabilities for long-horizon agents, continuous monitoring of web signals, event-driven architectures, and SQL-style programmable queries over the web. "We're building for what's next," the company said. Agrawal's tenure at Twitter ended in 2022 after Elon Musk's $44 billion takeover. With a background spanning IIT Bombay, Stanford, and a decade at Twitter, he is now focusing on his first major project since then, building Parallel Web Systems to advance AI infrastructure.

Former Twitter CEO Parag Agrawal emerges with AI startup

Parag Agrawal, the former Twitter CEO ousted by Elon Musk, has launched a new AI startup aimed at driving the next wave of AI agent research. Parag Agrawal, the former CEO of Twitter, now called X, has returned to the tech industry with a startup focused on helping artificial intelligence agents autonomously collect and analyze information from the web. Agrawal said in a LinkedIn post on Friday that his startup Parallel Web Systems Inc.'s first major product called Deep Research API outperforms humans and leading AI models -- including OpenAI's GPT-5 -- on some research benchmarks. "We already power millions of research tasks every day," Agrawal said, adding that a public firm is already automating human workflows with Parallel and coding agents are using its tools to find documents and debug issues. Agrawal has raised $30 million in funding for Parallel and leads a 25-person team, Bloomberg reported on Thursday. Parallel marks Agrawal's first big move in the tech space since Elon Musk ousted him and other C-level executives after the Tesla boss took over Twitter in October 2022, eventually changing the social media platform's name to X. The executives later sued Musk, claiming they were denied around $128 million in severance pay after their contracts were abruptly terminated. A district court ruled to move the case forward in November, but a final verdict hasn't been reached. Agrawal built machine learning models at Twitter while serving as chief technology officer under then-CEO Jack Dorsey. Agrawal then led Twitter from November 2021 until October 2022. After being fired from Twitter, Agrawal told Bloomberg that he dove back into AI research, reading papers and writing code again. He said he turned down several offers from struggling tech firms looking for him to "clean up shit" -- knowing that AI was what he wanted to focus on next. Agrawal's first AI idea was an AI-powered health-care startup, but he realized that AI agents, not humans, would become the biggest users of the web. Related: Crypto chases hype while missing foundational fortune "There'll be more agents on the internet than there are humans around. You will probably deploy 50 agents on your behalf to be on the internet," Agrawal told Bloomberg. "I think that's going to happen soon, like next year." Agrawal's bet on AI agents also lines up with recent predictions from Coinbase developers, who last week described autonomous AI agents as Ethereum's future "biggest power user." Coinbase's Kevin Leffew and Lincoln Murr said a little-used HTTP web standard, combined with Ethereum Improvement Proposal (EIP) 3009, could open a new era of e-commerce on the blockchain. AI agents could execute stablecoin transfers without human involvement -- enabling everything from self-driving taxis that pay for their fuel to AI models that monetize content on demand, the pair said.