VOYGR wants to fix the maps that keep lying to your AI

Try this experiment. Open your favorite AI assistant and ask it for a good dinner spot nearby. There's a decent chance it will hand you a restaurant that closed a year ago, moved across town, or now goes by a different name. The model isn't hallucinating, exactly. It's reading a map that stopped paying attention.

That gap between the world and the record of it is the whole business of VOYGR, a company from Y Combinator's Winter 2026 batch. Its pitch fits on an index card: maps tag places, VOYGR understands them. Underneath the slogan is a specific, unglamorous claim that the founders repeat without softening it - that Google Maps is wrong about roughly one in four businesses you look up. Not slightly outdated. Wrong.

For a person, a stale listing is a shrug and a second search. For an AI agent trying to book a table, route a courier, or match a credit-card transaction to a storefront, it's a failed action. As software starts doing things in the physical world rather than just describing it, the cost of a bad address stops being an annoyance and becomes a bug.

01 / The problemThe world moves faster than the map

The number VOYGR leans on is turnover: somewhere between 25 and 30 percent of places change every year. A cafe closes, an office relocates, a chain rebrands, a menu doubles in price. Traditional mapping data treats a place as a mostly static row - a name, a pin, a category, a phone number, maybe hours if you're lucky. VOYGR's argument is that a place is not a row. It's a thing that keeps happening.

If a fifth of what people ask AI models involves "where," "near me," or "is it open," then location is not a niche feature. It's a load-bearing input. And a smarter model stacked on top of stale data doesn't fix the problem - it just becomes more confidently wrong.

02 / The productTwo pipelines: check, then enrich

VOYGR splits the work in two. The first pipeline is validation: does this place still exist, and is it operating? Instead of trusting the incumbent maps, VOYGR cross-references the open web - news, articles, and public signals - and returns a verdict. The company reports a precision of 99.62% across its daily validations, and sorts results into a small, honest taxonomy.

That last chip is the tell of a serious data company. "Insufficient Data" is an answer most systems are too proud to give. VOYGR would rather say it doesn't know than pretend a ghost restaurant is still serving.

The second pipeline is enrichment. Where a standard mapping API might carry 10 to 15 attributes per place, VOYGR treats the profile as open-ended: foundational data (address, contacts, category), operating data (hours, amenities, menus, prices), and contextual data (reviews, news, events). The company likes to show off the range with a gag - it can tell you which sauna Justin Bieber frequents - but the serious version is that an agent can ask a place almost anything and get a fresh answer.

03 / The foundersThe people who built the old maps

The origin story is unusually literal. CEO Vlad Baskakov led product strategy and go-to-market for Google Maps' API products - bootstrapping new APIs, shaping how Maps and Gemini shared data, and working on the merchant side of the business. Before that he spent 15-plus years in growth and general-management roles across mapping, ridesharing and travel, with a stint at McKinsey and an MBA from Harvard. His co-founder and CTO, Yarik Markov, spent roughly two decades in tech and more than ten years leading infrastructure, machine-learning and search teams at Google, Meta and Apple.

In other words, they are not outsiders guessing at where mapping data breaks. They helped build the systems, watched the same failure modes for years, and left to build the version they wished existed - one where the customer is a machine, not a commuter.

04 / The wedgeSame price as Google Maps, none of the strings

VOYGR's commercial angle is quieter than its technical one, and possibly just as sharp. It prices roughly in line with Google Maps but drops the terms enterprises quietly resent - mandatory data-deletion windows and attribution requirements. If you're running millions of place records through an internal pipeline, "you're allowed to keep the data you paid for" turns out to be a real feature, not a footnote.

05 / The marketPlumbing for the agent era

The customer list VOYGR is chasing reads like a cross-section of the economy that runs on "where": mapping and geospatial companies, AI agent and LLM app builders, financial services matching transactions to storefronts, real-estate platforms, retail site selection, and advertising measurement. Early users and references include Overture Maps and RepRally. VOYGR is itself a member of the Overture Maps Foundation, which puts it inside the broader open-data movement trying to give AI a reliable picture of the physical world.

It's worth being clear-eyed about the risk, and VOYGR is. The obvious threat is that Google eventually loosens its own pricing and terms, and the incumbent advantage reasserts itself. The startup's answer is speed - lock in enterprise data pipelines before switching costs make the question moot. That's a race, not a moat. But the direction of travel is on VOYGR's side: the customer for location data is quietly shifting from humans who tolerate a wrong pin to agents that can't.

Whether VOYGR becomes the default layer or an acquisition target, the underlying observation will still be true in ten years: software that acts in the world needs data about the world that is actually current. Someone has to keep checking. VOYGR is betting that "someone" is a company, not a feature.