Ex-Twitter CEO Parag Agarwal returns with Rs. 260 crore AI startup that claims 10% better accuracy than GPT‑5

Parag Agrawal is making a big comeback just when many might expect him to take a breather. The Indian-origin technologist, and ex-CEO of Twitter, one of the world's most influential social media platforms, had his tenure cut short when Elon Musk took over Twitter.

But recently, instead of fading into the background, he’s resurfaced with a new venture that speaks to his vision for AI’s future by building tools that help intelligent systems truly understand and use the web, not just browse it.Parag Agrawal is returning to tech with his new AI startup, Parallel Web Systems Inc., based in Palo Alto. Founded in 2023, this venture has already drawn serious investor support, raising Rs 260 Crore from heavyweights like Khosla Ventures, Index Ventures, and First Round Capital, as reported by Bloomberg and TechCrunch.In a recent LinkedIn post, Agrawal described Parallel’s core mission as enabling companies to harness “web intelligence” for their products. He shared that Parallel’s platform already supports millions of daily research tasks for startups and enterprises. Agrawal also noted that one unnamed public company has integrated Parallel’s technology to automate workflows traditionally done by humans, achieving results that “exceed human-level accuracy”

"Some of the fastest‑growing AI companies use Parallel to bring web intelligence directly into their platform and agents.

A public company automates traditionally‑human workflows, exceeding human‑level accuracy with Parallel," he wrote in his LinkedIn post.

But what exactly is Parallel building?

The official website of the company describes itself as ‘The highest accuracy web search for your AI’ and ‘A Parallel web for AI’s. ’ According to the company, "AIs will use the web far more than humans ever have… We are creating systems and infrastructure for AIs to use the web effectively for completing complex tasks".This infrastructure includes a set of eight specialized “research engines”, each optimized for different speeds and depth of processing. The most advanced engine, named Ultra 8x, can spend up to 30 minutes poring over a query, and in benchmarks like BrowseComp and DeepResearch Bench, it beat both human researchers and OpenAI's GPT‑5 by more than 10 percentage points, as seen on their Official websites’ graphs.Parallel’s platform is already in action—processing millions of research tasks each day for startups and enterprises. The use cases are practical and diverse, as AI coding assistants pull live snippets from GitHub, retailers monitor competitors’ product listings, analysts compile customer reviews, and even automate workflows that were previously human-driven.