Alan Noon

500 million people installed Pokémon Go in 60 days. Every one of them became an unpaid 3D mapping contractor. Niantic introduced “AR Mapping” in 2020, framed as “Field Research.” Complete the scan, unlock a reward. Players walked circles around buildings, streetlights, and storefronts while their phone cameras captured geometry from every angle. The game awarded them a handful of Poké Balls. Niantic got geotagged photogrammetry data tagged with precise position, orientation, movement speed, and direction. 30 billion images. A million locations worldwide. Thousands of photos per location across different angles, weather conditions, lighting, and times of day. Building that dataset with paid contractors would cost tens of billions of dollars. Niantic built it for the cost of digital Pikachu rewards. Google figured this out a decade earlier with reCAPTCHA. “Click all the traffic lights” was never about proving you’re human. It was labeling training data for Street View, Maps, and autonomous vehicles. By 2011, reCAPTCHA users had digitized the entire Google Books archive and 13 million New York Times articles back to 1851. Google took a security product and turned billions of free human classifications into computer vision infrastructure. Niantic ran that same playbook at physical-world scale. The game was the CAPTCHA. Every AR scan was an unpaid mapping task dressed up as gameplay. Now the spin-out, Niantic Spatial, licenses centimeter-accurate visual positioning to Coco Robotics for delivery bots navigating city streets where GPS drifts 50+ meters. The system works because when a robot’s cameras see a building, Niantic already has thousands of photos of that exact building from every possible angle, tagged with sub-centimeter coordinates. No satellite required. Niantic raised $773 million and peaked at a $9 billion valuation. Last year they sold Pokémon Go to Scopely and kept the spatial data. The game generated revenue. The 30-billion-image 3D map of the physical world generated the company.

Escape from Butcher Bay’s menu cube is an impressive technical flex with clear roots in the European demoscene aesthetic (2004)

Inside of an arcade game manufacturing facility in 1982.

I recently came across this 3D NES emulator that transforms classic games into AR experiences. Seeing retro titles break out of the flat screen and coming to life in your physical space adds an entirely new layer of immersion and nostalgia.

Take It Down 'Omega Boost' PlayStation

Doing some more platform blending 😄 - Tripo initial highpoly generation - Hunyuan 3D Studio for lowpoly + auto UV unwrap - Trellis2 in ComfyUI for texturing + upscaling. Tripo worked the best with my 2 reference images (front side + side view) to create the initial 3D model. First time testing Hunyuan 3D Studio and it created the best lowpoly version with very good cylindrical shapes. The platform also has a UV unwrap tool so I don't have to do this manually in Blender. For texturing i'm using Trellis2 in ComfyUI. The projection is pretty solid and I can add nodes to upscale with AI to make the 2K textures more clean. Now the next step is rigging and animating to use in UEFN!