When Parag Agrawal was removed from Twitter’s headquarters at the end of 2022, the story line was one of the "end of an era." But for anybody acquainted with Agrawal as the Stanford-educated engineer who helped scale Twitter’s sophisticated AI systems, simply a pivot was undoubtedly in order. Flash forward to today, and Agrawal is now behind the launch of Parallel Web Systems a project that raises the expectation that his second act will be marked by the “plumbing” of the intelligence age.
From Chatbots to Infrastructure
While the world was distracted by the “poetry” of generative AI chatbots that can write sonnets or code Agrawal discovered a deeper rot in the system: The internet was made for humans, not agents. Today’s AI models often “hallucinate” because they literally are attempting to make their way in a world of CAPTCHAs, paywalls, unstructured data armed with devices crafted for thumb and screen.
Parallel Web Systems is Agrawal’s response to that friction. Rather than constructing a second model, he is constructing a “machine-first internet.”
What is Parallel Web Systems?
Parallel is a layer of infrastructure that is designed to give AI agents their own “browser.” It offers a set of APIs to allow AI systems to search, validate and organize real-time web data at a degree of granularity that large language models (LLMs) cannot do for themselves.
The Performance Gap: As recent benchmarks have shown, Parallel’s research engines have also been reported to outperform human researchers and high-level models such as GPT-5 in the extraction of deep-web data.
The "Ultra" Engines: There are eight specialized research engines, which have come to be known as the “Ultra” engines that perform anything from high-speed “Ultra1x” (60-second results) to “Ultra8x,” which can spend 30 minutes diving deep for verifiable research.
The $130 Million Bet on Stability
The venture community has already signaled its agreement with Agrawal’s “infrastructure-first” thesis. Parallel just closed a $100 million Series A round (after a $30 million seed round), backed by heavyweight investors such as Kleiner Perkins and Index Ventures.
This valuation, about $740 million, isn't driven by clout in social media; it’s based on the value of a "plumbing" job. Agrawal’s group of engineers, spanning veterans at Google, Stripe and Airbnb, is building an "open market mechanism" designed to encourage publishers to keep their content available to AI agents for fair compensation solving for the future and potentially ending the conflicts between copyright makers and competitors that plague the industry.
The Second Act
Agrawal’s time at Twitter often came as a shadow against the noise of the public square. At Parallel, he has once again reentered the shadows, the places he is probably most dangerous: at the architectural level. If his first act was simply to manage the chaos of human conversation, his second is to arrange the logic of machine intelligence.
By creating the rails that will make AI trustworthy rather than simply impressive, Agrawal has the potential to leave a legacy that is much longer than a 280-character tweet.
