Parag Agrawal, the former Twitter CEO ousted after Elon Musk's 2022 takeover, has reemerged in Silicon Valley with Parallel Web Systems, a startup promising to push artificial intelligence beyond current limits.
The company raised $30 million from Khosla Ventures, Index Ventures, and First Round Capital to launch its flagship product, the Deep Research application programming interface, Entrepreneur reports.
Deep Research API Targets Benchmarks Against GPT-5
Parallel's Deep Research API already handles millions of requests daily, ranging from document discovery to code debugging and workflow automation. According to Entrepreneur, the system integrates eight research engines designed for specific functions like long-form synthesis, cross-disciplinary analysis, and knowledge retrieval.
Don't Miss:
- The same firms that backed Uber, Venmo and eBay are investing in this pre-IPO company disrupting a $1.8T market — and you can too at just $2.90/share.
- Kevin O'Leary Says Real Estate's Been a Smart Bet for 200 Years — This Platform Lets Anyone Tap Into It
Benchmark tests published by the company claim the system has already outperformed OpenAI's GPT-5 on research-oriented tasks that require multi-step reasoning and contextual synthesis.
On BrowseComp, which measures how well systems navigate the live web to answer open-ended questions, Parallel reached 58% accuracy compared with GPT-5's 41% and a human baseline of 25%. On DeepResearch Bench, which evaluates long-form research tasks across 22 fields, Parallel achieved an 82% win rate against the reference standard, surpassing GPT-5's 66%.
Parallel Web Systems attributes this performance to infrastructure built from the ground up for AI agents, delivering structured, verifiable, and real-time results.
Independent reviews remain limited, but Entrepreneur says demonstrations show the product can manage complex queries, execute multi-hop reasoning, and generate outputs with factual provenance.
Trending: 7 Million Gamers Already Trust Gameflip With Their Digital Assets — Now You Can Own a Stake in the Platform
A ‘Programmatic Web' Built for AI Agents
Palo Alto-based Parallel Web Systems describes its mission as preparing for a shift where AI agents become the primary users of the web. "AIs will use the web far more than humans ever have," the company wrote in the announcement.
The firm advocates for a "programmatic web" designed specifically for machine consumption, built with declarative interfaces, transparent attribution, and value-driven open markets where sources and contributors are directly credited. Parallel Web Systems argues that the existing web, optimized for human clicks and ad impressions, cannot adequately serve AI systems that need scalable access to factual information.
Potential applications of this model extend to legal analysis, market intelligence, and scientific research, where Entrepreneur says autonomous systems need reliable access to live data rather than static training sets. Parallel emphasizes system and organization controls 2 Type 2 certification, and enterprise readiness to meet developer demands across industries.
See Also: 2,000 High Earners Manage $6B With This AI Platform — Learn More
Agrawal's Second Act in Tech
Agrawal's exit from Twitter came after months of conflict with Musk. In 2022, Musk accused Twitter of misleading shareholders and tried to walk away from his $44 billion acquisition. Twitter's board sued Musk to enforce the deal, a legal battle that Agrawal was forced to fight while still leading the company.
The first thing Musk did after closing the purchase was fire most of the executive team, including Agrawal, without severance. Friends and colleagues urged Agrawal to take a break, but Bloomberg says he instead returned to Palo Alto, where he spent his time reading research papers, writing code, and considering what to build next.
Agrawal views Parallel as a chance to define himself beyond Twitter, telling Bloomberg, "I don't think Twitter defines me. If I do a good job, I'm hoping this company will define me."
Read Next: Shaquille O'Neal Wants People to Take Heart Health Seriously — This AI-ECG Could Make That Easier
Image: Shutterstock
