Skip to content

By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.

Sequoia CapitalSequoia Capital

Parallel’s Parag Agrawal: Building a New Web for AI Agents

Parag Agrawal is making a bet that goes against two decades of web search: agents will query the web a thousand times more than humans ever have, and the infrastructure built around human clicks is wrong for them.

Parag AgrawalguestSonya HuanghostAndrew Reedhost
Aug 25, 202655mWatch on YouTube ↗

Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.

At a glance

WHAT IT’S REALLY ABOUT

Parallel builds agent-first web search and new incentives for content

  1. Parallel is building search infrastructure and products designed specifically for AI agents to search and use the web at far higher volumes than humans.
  2. Parag Agrawal explains web search as crawling, indexing, and multi-stage retrieval/ranking to match massive numbers of pages to massive numbers of queries, then reframes it as selecting the highest-signal tokens under strict latency budgets.
  3. Parallel chose to launch a search-agent product before a full index, using longer inference-time crawling to serve real workflows and incrementally build a broader index over time.
  4. He argues agentic search differs from human search because agents can issue richer queries and consume authoritative excerpts directly, reducing reliance on human click-based signals and human-oriented SEO content.
  5. Agrawal contends the shift from human visits to agent access disrupts ad-based monetization, motivating new incentive-aligned pricing and attribution mechanisms for web content, including Shapley-value-inspired approaches.

IDEAS WORTH REMEMBERING

5 ideas

Parallel treats human click data as the wrong feedback loop for agentic search.

He argues that click data reflects human constraints (like impatience and interface friction) rather than what an agent should learn from, and that agent-generated feedback plus model-based compression can make search indexing/ranking more tractable for a new entrant.

Parallel started with search agents first to bootstrap toward a full index.

Parallel launched a search-agent product that could crawl after a query arrived (including minute-long or even longer “deep research”), using latency tolerance to compensate for not having a full index on day one while they built coverage over time.

Agentic search is an optimization problem across quality, cost, and latency.

He frames the core search job as selecting the “right 1,000 tokens” from an enormous web corpus under tight latency budgets, describing a multi-index, multi-stage retrieval and ranking pipeline with different compute tradeoffs across product tiers.

Serving agents changes the search interface and query formulation.

He says agents can provide longer, clearer queries than humans (who tend to type short, typo-filled keywords), allowing the system to rely less on guessing intent and more on well-specified requests.

Agents can bypass human web “slop” by pulling precise excerpts from authoritative sources.

He explains that some “SEO” content exists because humans avoid slow authoritative sources (like navigating long PDFs), while agentic search can extract the authoritative excerpt and deliver it directly, avoiding the human-centric tradeoff between authority and convenience.

WORDS WORTH SAVING

5 quotes

Our view at Parallel is that human click data is a bug. An agent doing work with search should rely on agent feedback, not human feedback.

— Parag Agrawal

We started Parallel with the bet that agents would do it 1,000X more than humans ever have.

— Parag Agrawal

What we build is something that multiplies on top of a model in order to give either you can call it the model m- gets better or the agent built with the model gets better and has superpowers, right?

— Parag Agrawal

It's at its core an optimization problem around quality, cost, and latency.

— Parag Agrawal

If we don't figure out a new business model, we're seeing it already, right? Like, people are gonna say, "Okay, I don't want my content to be accessed by an agent because I have a business model for humans, so I actually wanna go pay someone to SEO out, optimize myself, so more humans show up. But then if their agent shows up, I'm gonna cut it off."

— Parag Agrawal

Agentic search versus human searchCrawling, indexing, retrieval, and ranking pipelineQuality–cost–latency tradeoffs and “Turbo”Search agents bootstrapping index coverageAgents handling authoritative sources vs SEO pagesPartnership integration as search/grounding providerWeb monetization shift beyond ads and content attribution via Shapley values

SPEAKERS

  • Parag Agrawal

    guest

    Founder of Parallel and former Twitter CEO focused on building agent-first web search and infrastructure.

  • Sonya Huang

    host

    Sequoia Capital host of the Training Data podcast.

  • Andrew Reed

    host

    Sequoia Capital co-host on Training Data who joins the conversation with questions and prompts.

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

Read along on your phone, search it beside the video, or ask your favorite AI app what was said.

iOSAndroidClaudeChatGPTChrome