https://youtu.be/iAjccA_HH9M?si=UVBjI89EuhQvlsrI
Parag Agrawal — the former Twitter CEO — has resurfaced in a rather interesting place: he is now founder and CEO of Parallel Web Systems, an AI infrastructure company building the web layer for AI agents rather than for human browsing. As of September 2026, Parallel is valued at about $2 billion after raising a $100 million Series B led by Sequoia; total funding is around $230 million.
The central idea is quite different from creating another ChatGPT. Parallel builds search and research infrastructure that lets autonomous agents go out onto the live web, retrieve information, reason over many sources, and return structured results. Its customers reportedly include Harvey, Notion, Clay and Opendoor, and Parallel says more than 100,000 developers are using its products.
More interestingly, in July 2026 Parallel deepened its partnership with Google Cloud: Google is making Parallel’s search/retrieval tools available to customers building agents with Gemini. So Agrawal is effectively betting on something like:
LLM → agent → Parallel → open web
rather than
human → Google → websites.
There is another clever aspect. Parallel launched an initiative called Index, intended to compensate publishers/content owners when AI agents use their material. Agrawal has been arguing that the emerging “agentic web” needs an economic architecture in which machines can retrieve web content without simply destroying the incentives to produce it.
So his trajectory after Musk fired him from Twitter in October 2022 is rather striking. He has moved down one layer in the stack:
Twitter: What should humans read?
Parallel: How should machines read the entire web?
And I think the second problem may ultimately be much larger.
It also connects rather neatly with our recent discussion about Manus and agents. Manus, ChatGPT agents, Gemini agents, Harvey, etc. all eventually encounter the same bottleneck: the model may be intelligent, but it needs a reliable sensory organ for the changing external world. Parallel is trying to become that sensory organ — almost an API-level Google built specifically for nonhuman readers.
That makes Agrawal substantially more interesting today than he was as Twitter CEO. He is no longer fighting over a social network; he is competing for a piece of the infrastructure of the agentic Internet.
I can also compare Parallel vs Google Search vs Perplexity vs OpenAI browsing/search — because they occupy surprisingly different layers of the emerging AI stack.
