Executive Overview
For nearly three decades, the architecture of the internet has been meticulously engineered for the human eye. Traditional search engines were forged to cater to users who lack the time, patience, or cognitive bandwidth to manually scan through sprawling, text-heavy web pages. They delivered the quintessential "ten blue links"—a format that optimized keyword matching and human browsing behavior.
However, the rapid ascent of generative artificial intelligence and autonomous AI agents has fundamentally broken this paradigm. Unlike humans, modern large language models (LLMs) and autonomous agents can ingest, parse, and synthesize vast oceans of information in milliseconds. They do not need summaries designed for casual scrollers; they require deep, unfiltered access to raw data, hyper-efficient indexing, and real-time grounding sources to perform complex multi-step reasoning.
Enter Keenable, a stealth-born startup aiming to solve this infrastructural mismatch. Co-founded by former Yandex search chief Andrey Styskin and German AI scientist Matthias Petri, Keenable has officially emerged from stealth with a formidable $26 million in seed funding. The round was led by venture capital firm Accel, with additional participation from Conviction Partners and prominent angel investors.
Keenable is not building a consumer-facing search engine to rival Google. Instead, it is constructing a massive, web-scale search infrastructure tailor-made for AI models, agents, and inference providers. Boasting an active index of over 100 billion documents—and already deployed in production environments by leading AI labs—Keenable represents the vanguard of a broader industry shift: the quiet death of the traditional search engine and the birth of a web infrastructure engineered explicitly for machines.
Detailed Chronology: From Yandex and Amazon to the Birth of Keenable
The intellectual origins of Keenable trace back years before the company’s formal incorporation, forged in the trenches of some of the world’s largest technology enterprises.
The Seeds of an Idea
Andrey Styskin spent two decades architecting search capabilities, most notably leading Russian search giant Yandex’s core search, AI, and cloud divisions. Following his tenure at Yandex, Styskin transitioned to Amazon, where he collaborated with German AI scientist Matthias Petri on advanced web search infrastructure for consumer AI applications, most notably Alexa.
During their time at Amazon and observing broader industry shifts, Styskin and Petri noticed a profound architectural bottleneck. Traditional enterprise search solutions, while effective for enclosed corporate intranets, catastrophically break down and rack up unsustainable financial costs when scaled to the entire open internet. Simultaneously, data from infrastructure providers like Cloudflare revealed an explosive surge in web traffic driven not by humans, but by automated AI scrapers and crawlers—including OpenAI’s GPTBot and Anthropic’s ClaudeBot.
Styskin realized that the burgeoning agentic AI ecosystem was operating on legacy plumbing. Existing consumer search engines were designed to filter out information to save human time, whereas AI systems suffered precisely because they lacked low-latency, low-cost access to exhaustive, real-world source material.
Stepping Out of Stealth
Recognizing the massive gap in the market, Styskin tapped into his deep industry network to recruit a lean, elite squad of former colleagues and engineering talent. Keenable was quietly established to build proprietary retrieval capabilities optimized for machine consumption.
The startup’s development accelerated rapidly, culminating in a successful $26 million seed funding round led by Accel. Zhenya Loginov, the Accel partner who spearheaded the investment, saw an acute vulnerability in the current tech landscape. With tech giants tightening their grip on data and actively phasing out public access to their search APIs, AI startups found themselves starved of reliable web-scale infrastructure options.
Keenable seized this vacuum, scaling its web index past 100 billion documents and securing early traction. The startup’s API is already integrated into the production pipelines of several undisclosed AI labs and inference providers, utilized during both model training and runtime inference. Furthermore, Keenable recently announced a high-profile partnership with voice AI pioneer Gradium to power real-time, low-latency information retrieval for conversational voice agents.
Supporting Context & Metrics: The Mechanics of Machine-First Search
To understand why Keenable commands a $26 million seed valuation, one must examine the staggering economic and computational realities of web-scale AI search.
The Economics of Exhaustive Scraping
Building and maintaining a web-scale search index is, in Styskin’s own blunt assessment, "painfully expensive." The sheer volume of the modern internet—compounded by video transcripts, deep forum threads, academic papers, and dynamic web applications—presents a monumental data processing challenge.
If an AI application attempts to query the broader web without a specialized, fine-tuned index structure, the computational overhead required to scan and serve the entire internet becomes economically unviable. Styskin explains Keenable’s core technological differentiator:
"If you do not fine-tune your index structures for a specific task, the cost of serving and scanning the whole internet is enormous because of the volume. That’s why you need to innovate on how you can narrow the search space based on your query very fast. This is what we are bringing to the table."
Breaking the Tech Giant Monopoly
For years, developers building AI applications relied heavily on established search APIs provided by Google and Microsoft (Bing). However, facing the economic pressures of feature cannibalization and monetization shifts, these tech giants have systematically altered their strategies. Google and Microsoft have begun scaling back or shutting down their legacy custom search APIs, opting instead for heavily bundled, closed ecosystems where they exercise absolute control over partnerships.
This insular approach created a severe supply crunch for independent AI labs, agent developers, and inference providers. Keenable stepped directly into this void. By offering an open, highly specialized, and cost-efficient API, Keenable allows AI companies to bypass the walled gardens of Silicon Valley.
Proprietary Architecture: The Web Query Language
Beyond raw indexing, Keenable is pushing the envelope with proprietary retrieval capabilities. The startup is developing an upcoming product tentatively titled the Web Query Language (WQL).
Unlike traditional keyword-based queries, WQL is engineered to assist AI systems in complex, multi-hop reasoning tasks. If an AI agent is asked a complex question requiring data points scattered across five different web domains—none of which individually contain the complete answer—WQL orchestrates the retrieval, synthesis, and alignment of these disparate sources natively, presenting a cohesive dataset for the agent to reason over.
Official Statements & Industry Perspectives
The structural transition from human-centric to agent-centric search has ignited intense debate across the venture capital and AI research communities.
Andrey Styskin, Co-Founder of Keenable:
Styskin is acutely aware of the David-versus-Goliath dynamic facing his startup. Challenging incumbents like Google in search is widely considered a fool’s errand. However, Styskin points to the classic business theory of The Innovator’s Dilemma to explain why Google is uniquely vulnerable in the age of agentic queries. Because Google’s revenue model is inextricably tied to human browsing habits, display ads, and the traditional search engine results page (SERP), the tech titan struggles to pivot its core infrastructure toward autonomous AI agents without undermining its own cash cow.
"This actually creates a new flywheel that is different from what Google learned from human behavior," Styskin noted in an interview with TechCrunch. He believes that a nimble, specialized startup can innovate faster and deliver a far more cost-efficient retrieval solution for the next generation of AI enterprises.
Zhenya Loginov, Accel Partner:
Loginov, who led Accel’s investment in Keenable, emphasized the scarcity of infrastructure options currently available to independent AI developers. With legacy APIs vanishing, Loginov views Keenable not merely as an alternative search tool, but as vital foundational plumbing for the entire artificial intelligence economy.
"AI players have very few options when it comes to web-scale search infrastructure," Loginov noted, underscoring the urgency that prompted Accel to back Keenable’s vision at such an early stage.
Future Outlook: The Post-Ten-Blue-Links Era
As Keenable transitions out of stealth, its immediate roadmap is clear. The company currently operates with a lean, highly specialized engineering team of 15 distributed across the United States and Europe. Armed with its $26 million capital infusion, Keenable plans to double its headcount by the end of the year, aggressively scaling its go-to-market operations, expanding its engineering footprint, and onboarding new enterprise clients.
Yet, Keenable is far from alone in recognizing this monumental shift. The race to build the ultimate machine-readable search layer is heating up. Competitors such as Brave (via its dedicated search API) and Exa.ai (which specializes in neural search for LLMs) are actively vying for supremacy in the AI data retrieval market. Meanwhile, Google itself is desperately attempting to reinvent its legacy search experience to survive the generative AI wave.
Despite the crowded landscape and the immense financial hurdles of maintaining a 100-billion-document index, Keenable’s core thesis resonates deeply with the trajectory of modern computing. Whether the end-user is a human browsing on a smartphone or an autonomous agent executing complex financial workflows across enterprise networks, one reality has become definitively clear:
The era of the "ten blue links" is rapidly coming to a close. The future belongs to those who can feed the machine.
