The editorial argues that while privacy concerns are valid, the more significant insight is the economic breakthrough: Niantic gamified data collection so effectively that 10M+ scans across a million locations were captured for free, producing dense multi-angle imagery that no satellite operator or AV company has been able to assemble at any price. This redefines the cost structure of building geospatial foundation models.
The DroneXL report frames Vantor's DoD partnership as part of the broader 'assured PNT' scramble triggered by Ukraine's daily GPS jamming reality. The Pentagon's interest is explicitly framed as preparing for a Pacific theater where satellite navigation is assumed denied, putting Vantor in direct competition with Sandboxx and Shield AI for the visual-positioning niche.
The editorial notes Niantic 'quietly' turned players into a sensor network over nearly a decade, with the Wayspot scan mechanic disguised as a reward loop rather than disclosed as military-grade geospatial data collection. The framing — players earning in-game rewards while their phones captured IMU-tagged video of alleys and building facades — implies the consent model never contemplated downstream defense applications.
Niantic, the company behind Pokémon Go, Ingress, and Pikmin Bloom, has spent nearly a decade quietly turning its players into a global, unpaid sensor network. In 2024 it spun the resulting dataset and ML stack into a separate entity called Vantor, pitched as a 'large geospatial model' company. As of this week, Vantor's flagship customer is no longer a tourism app or an AR headset vendor — it's the U.S. Department of Defense, via a partnership to provide GPS-denied visual positioning for military drones.
The mechanic is simple and, in hindsight, obvious. Every time a Pokémon Go player completed a 'Wayspot scan' — slowly panning their phone around a PokéStop to earn rewards — Niantic captured a short video plus IMU data, geotagged to within a few meters. Multiply that by a reported 10 million+ scans across more than a million locations, and you have something no satellite operator and no autonomous-vehicle company has: dense, ground-level, multi-angle imagery of the exact street furniture, alleys, and building facades that GPS struggles with. Vantor's model takes a single camera frame from a moving drone and matches it back to that scan corpus to produce a position fix, no satellite required.
The DroneXL report names Anduril-adjacent integrators as early evaluators, and confirms Vantor is targeting the same 'assured PNT' (positioning, navigation, timing) niche that companies like Sandboxx and Shield AI have been scrambling to fill since Ukraine made GPS jamming a daily reality. Niantic's executives have framed this as 'spatial computing for the physical world.' The Pentagon is framing it as a hedge against a Pacific theater where GPS is assumed compromised on day one.
The story everyone is telling on Hacker News is the privacy one, and it's not wrong — but it's the second-most-interesting thing here. The first is that Niantic just demonstrated the cheapest known method of collecting high-resolution, ground-truth geospatial data: dress the labeling task up as a mini-game and let users pay you for the privilege. Tesla pays for Autopilot miles by selling cars. Mobileye pays Mobileye-equipped fleets. Niantic paid nothing. Players opted in voluntarily, in exchange for a Pokéball.
Compare the unit economics. A LiDAR-equipped survey vehicle costs roughly $400–800 per mile of road mapped. Commercial drone photogrammetry runs $100–300 per acre. A single Wayspot scan — 10 seconds of phone video plus IMU — costs Niantic effectively the marginal compute of validating it, call it fractions of a cent. At 10M+ scans the implied dataset would have cost a defense contractor on the order of $1B+ to assemble conventionally. Niantic assembled it for the cost of a server farm and some Pikachu artwork.
The second thing worth noting is the technical fit. GPS-denied navigation has historically used inertial drift correction, terrain contour matching (TERCOM, from cruise missile days), or expensive star trackers. None of those work well at low altitude in cities, which is exactly where small drones now operate. Visual place recognition — 'I've seen this corner before, and I know where it is' — has been a research problem for 20 years. The bottleneck was never the algorithm; it was the absence of a dense, labeled, ground-level reference corpus. Pokémon Go quietly solved that.
The community reaction has split predictably. Privacy-focused HN commenters point out that Niantic's terms of service have, since at least 2019, granted them a 'perpetual, irrevocable, worldwide, royalty-free' license to use scan data. Defense-tech commenters point out, correctly, that this is one of the few cases where a U.S. company actually has a defensible moat against PRC competitors who can't legally operate Pokémon Go in China. Both are right. The interesting question is whether players, told plainly that their walks were training drone models, would have kept playing. The empirical answer, based on every prior consumer-data scandal: yes, almost certainly.
If you build consumer apps with any sensor surface — camera, GPS, microphone, IMU, BLE — Vantor is now the proof-of-concept that your gameplay loop is potentially a training dataset for a much larger market. The practical implications:
Audit your ToS for downstream rights. If your license grant is narrowly scoped to 'providing the service,' you cannot do what Niantic did. If it's the standard 'perpetual, worldwide, sublicensable,' you can. Most app boilerplate ToS templates default to the broad form. Most founders have never read theirs carefully enough to know.
Think about which sensor signals have a B2B buyer. Camera-plus-IMU has obvious geospatial value. Microphone arrays have acoustic-localization value (drone detection, gunshot detection). BLE scan logs have indoor-positioning value. The naive consumer-app monetization is ads; the Niantic monetization is selling the byproduct of usage to someone who'd otherwise spend nine figures collecting it themselves.
Assume your competitors will do this whether you do or not. If you're building a fitness app, a navigation app, an AR try-on app — the dataset you're sitting on is worth more than your subscription revenue. Strava already learned this the hard way in 2018 when its heatmap inadvertently exposed military base locations. Strava's mistake was giving the data away; Niantic's innovation is selling it.
The DoD contract is almost certainly the first of several. Expect Vantor to expand into autonomous-vehicle reference maps, AR cloud anchors for headset makers, and — most lucratively — last-mile delivery drones, where GPS multipath in cities is a more boring but much bigger market than military jamming. The broader pattern to watch is the inversion of consumer apps into infrastructure: the game was never the product, the scans were. Expect a wave of similar spin-outs over the next 18 months as every consumer-sensor company with a board to please discovers their dataset has a second customer. Whether players, regulators, or your own engineering team are comfortable with that is a different question — but the economics are now public, and they are extremely hard to ignore.
If you are looking for something to channel that energy into, you could help improve open street map using streetcomplete: https://streetcomplete.app/
I stopped scanning pokestops because the effort has outweighed the rewards. A lot of the time, the requests show up as "research tasks" for a point of interest that I quickly passed by and have no interest in returning to, besides the tasks related to taking pictures of your buddy pokemon
This article is based on reporting from Trouw: https://www.trouw.nl/redactie/PokemonGo/I was interviewed for the Trouw piece and briefly quoted. This isn't to detract from the DroneXL piece, which adds its own angle.
This shouldn't be a surprise. But at this point it feels like if you don't completely avoid participating in digital society, your data will be used against you or groups/countries you support.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
As someone who works in this space, the headline is a bit of a stretch. The overlap in the locations of Pokemon Go Player data and any active Drone heavy theaters of war is a tiny sliver (or zero?).The military contractor (Vantar/Maxar) in question basically admits so but just "reserves th