Pokémon Go 背后的人：靠美国中情局投资起家，如今正替军工 AI 做全球测绘
Pokémon Go 背后的人：靠美国中情局投资起家，如今正替军工 AI 做全球测绘
Human history has a few moments when mass-scale geospatial data collection quietly rewired power.
In the Age of Sail, Portugal and Spain funded fleets to draw sea charts; whoever owned the most accurate routes owned trade and colonial leverage. In World War II, military mapping became a decisive operational advantage—later leaders openly credited maps as a critical ingredient of victory. During the Cold War, spy satellites turned territory into pixels, and analysts spent their days counting missile silos from blurry imagery.
In 2026, we are watching a new chapter form: a consumer game engine that turned into a planetary sensing network, and a new generation of “AI maps” designed not just for humans to look at, but for machines to act on.
And that raises a question that the crypto industry cannot ignore:
When the world is being re-mapped for AI—and increasingly for defense—who owns the data, who can audit it, and who gets paid?
From Keyhole to Google Earth: the “CIA-funded” origin story (and what it really means)
The person behind this story is John Hanke, best known to the public as the founder of Niantic and the driving force behind Pokémon Go.
Before Niantic, Hanke led Keyhole, a company building interactive 3D Earth visualization—technology that would later become part of Google’s geospatial stack. What makes Keyhole historically significant is how early national security interest converged with commercial mapping.
In 2003, In-Q-Tel—a venture organization funded by the U.S. intelligence community—announced a strategic investment in Keyhole, noting the investment was made in February 2003 and tied to the needs of the then National Imagery and Mapping Agency (NIMA). You can read the original announcement in the In-Q-Tel press release. A later deep dive on the intelligence-to-consumer pipeline is also captured in The Guardian’s reporting on Keyhole and Google Earth.
This isn’t a conspiracy trope. It’s a recurring pattern in tech history:
- A government problem (situational awareness) creates early funding.
- A commercial platform scales the interface and distribution.
- The result becomes a dual-use infrastructure layer.
Crypto people should recognize the structure immediately—because blockchains are also dual-use infrastructure. The difference is that blockchains can be designed to make power more legible and contestable.
Pokémon Go as a mapping machine: “play” was the UX for sensor deployment
Pokémon Go looked like a game about catching creatures. Under the hood, it normalized a behavior that is priceless for mapping:
- walking to specific coordinates,
- scanning landmarks,
- submitting points-of-interest,
- and continuously validating real-world locations through repeated human presence.
Today Niantic’s mapping efforts are explicit. Niantic Spatial documents ongoing collection of geospatial data and claims privacy measures intended to reduce accidental collection of personal data in its Niantic Map Data Collection policy page.
Then came the strategic pivot: Niantic moved to split its consumer games business from its mapping / platform direction, positioning the new entity as a dedicated geospatial AI company. Niantic publicly described this restructuring in March 2025 in Evolving Niantic Spatial Inc..
So the story arc is no longer “game studio builds AR.” It’s:
consumer engagement → global scanning → large-scale spatial model → enterprise and public sector deployment
The new prize: “AI-powered maps” for machines, not just humans
Traditional maps answer: Where am I? What’s nearby?
The new generation of mapping answers: What am I looking at, and how do I operate here? That requires machine-perceivable geometry, semantics, and localization.
Niantic Spatial is marketing a stack that includes visual positioning designed to work even when GPS is unreliable—language that directly overlaps with defense requirements. For example, Niantic Spatial’s product materials for localization emphasize positioning “anywhere in the world,” including GPS-denied settings on its VPS / Localize pages. The company also publishes a dedicated overview for government and defense contexts on its Public Sector and Defense page.
And this is not purely theoretical positioning. In 2025, an Aviation Week analysis of “Project Orbion” described Niantic Spatial providing large geospatial model reconstruction and visualization services, with the resulting capability aimed at use cases including disaster monitoring and—explicitly—tracking troop movements and shipping routes; it also noted a U.S. Coast Guard training center test case (PDF excerpt hosted by Aechelon).
Separately, Niantic Spatial has publicly announced a multi-year partnership with Snap to build a shared AI-powered map of the real world (Niantic Spatial announcement, June 10 2025).
If you work in crypto, you should translate all of this into one sentence:
Geospatial data is becoming the training set and operating system for embodied AI.
And whoever controls that layer can set rules for access, pricing, censorship, and surveillance.
Where crypto enters: geospatial data needs provenance, incentives, and privacy
Blockchain is not a magic shield against militarization. But it is uniquely good at three things that geospatial AI desperately needs:
1) Provenance: “Who captured this, when, and under what conditions?”
AI mapping systems are only as reliable as their inputs. But in the real world, inputs are messy:
- spoofed GPS,
- tampered imagery,
- synthetic uploads,
- incentives to game rewards,
- and political incentives to falsify territory.
A properly designed onchain pipeline can create an auditable trail for:
- capture metadata commitments,
- device attestations (where feasible),
- validation signatures,
- and dispute resolution.
This matters for consumer applications (AR games, delivery, robotics), but it matters even more when data influences public safety or national security decisions.
2) Incentives: paying for coverage without building a monopoly
One reason centralized mapping wins is simple: it’s expensive. Crypto introduced a new template—token incentives for decentralized physical infrastructure networks (DePIN)—that can fund coverage without a single company owning all sensors.
By 2025, DePIN became a mainstream narrative inside Web3 conferences and media, often framed as “the eyes and ears of AI.” A quick snapshot of that discourse is captured in Forbes’ ETHDenver 2025 trend report, including emphasis on proof systems that resist location spoofing.
The deeper point isn’t hype. It’s industrial organization:
- Centralized model: one company funds capture → owns dataset → sells access.
- Open network model: many contributors capture → protocol coordinates verification → value accrues to network participants.
3) Privacy: minimizing “mapping = surveillance”
The biggest user concern is not whether maps exist; it’s whether mapping becomes a default surveillance layer.
Crypto primitives can help, if used honestly:
- zero-knowledge proofs for “I was at an allowed location” without revealing the full trail,
- encrypted data marketplaces where buyers can run queries without getting raw footage,
- and user-controlled permissions that are enforceable by keys, not by terms-of-service.
If geospatial AI is the new strategic resource, then privacy-preserving verification is the new civil liberty battleground.
The uncomfortable reality: mapping networks are inherently dual-use
It’s tempting to frame this as “games vs military.” In practice, mapping technology rarely stays in one lane.
- A feature that enables accurate AR anchoring also enables precise navigation for autonomous systems.
- A dataset that helps disaster response can also help targeting.
- A localization stack built for shopping malls can be repurposed for GPS-denied environments.
That is exactly why the crypto question is not “can we stop dual-use?”
It’s:
Can we build neutral infrastructure where contributors retain agency, where data access is transparent, and where value doesn’t concentrate into a single black box?
A practical design checklist for “onchain mapping” in 2025–2026
If you’re building or contributing to decentralized mapping or proof of location systems, users now demand answers to these questions:
- What exactly is collected? (raw images, point clouds, features, derived embeddings)
- What is stored offchain vs onchain? (hash commitments onchain; heavy data offchain)
- Who can buy access, and can access be revoked?
- How do you mitigate spoofing? (multi-sensor fusion, challenge-response, reputation, staking + slashing)
- What is the exit right? (can contributors delete, rotate keys, or stop future licensing?)
- How do you handle compliance without becoming a surveillance honeypot?
Niantic’s own public mapping disclosures show the direction of travel: large-scale capture, privacy language, and global coverage ambitions (Niantic Map Data Collection). Crypto projects attempting similar capture must be even more explicit—because Web3 users are rightly skeptical.
Self-custody becomes operational security when your wallet is tied to real-world data
There’s an overlooked shift happening:
In DePIN and geospatial networks, a wallet is no longer just an investment account. It becomes:
- your contributor identity,
- your rewards endpoint,
- your governance voice,
- and sometimes the authorization key for data uploads or validator roles.
That makes self-custody not merely philosophical, but practical operational security.
If you are earning tokens by mapping, scanning, or operating devices in the real world, you are exposed to:
- phishing for approvals and signatures,
- malware that swaps payout addresses,
- SIM-swap recovery attacks on hosted wallets,
- and social engineering aimed at draining contributor accounts.
A hardware wallet helps by keeping private keys offline, so “approve” cannot be silently triggered by a compromised laptop or browser session.
If you want a clean setup, OneKey hardware wallets are designed for self-custody workflows: keep keys offline, review transactions on-device, and separate higher-risk “daily” wallets from long-term holdings. In a world where geospatial data is being financialized—and sometimes politicized—that separation is not paranoia; it’s basic hygiene.
Closing: the next map war is about verification
The Age of Sail rewarded whoever drew the best charts. The Cold War rewarded whoever captured the best satellite imagery. The AI era will reward whoever owns the most useful machine-readable model of reality.
John Hanke’s path—from a Keyhole investment tied to U.S. intelligence needs (In-Q-Tel’s 2003 announcement) to consumer-scale scanning and now to enterprise and defense-facing geospatial AI (Niantic Spatial public sector materials, plus reporting on “Project Orbion” in this Aviation Week PDF excerpt)—is not an anomaly. It’s a preview.
For the crypto industry, the strategic opportunity is clear:
Build mapping and location networks where data provenance is auditable, incentives are distributed, and privacy is enforceable by cryptography—not by promises.
Because in the next decade, the most important question won’t be “who has the map?”
It will be:
who can prove it’s true—and who can prove you still have a choice.
