Japanese braille shows up in places most people walk past without thinking — the textured rail at a train station, the edge of a pill box, the panel beside an elevator button. It's present, it carries information, and almost nobody can read it. TenjiScan started from that observation. Point your iPhone camera at Japanese braille and get the text back on the spot. That's the whole pitch. Why on-device processing mattered The obvious implementation would be: send the photo to a server, run recognition there, return text. Faster to ship, potentially higher accuracy. We didn't do that, for two reasons. First, the places where you'd actually want to read braille — subway stations, hospital hallways — are often places with weak or no connectivity. Second, you're sometimes photographing medication labels or personal items. Sending those images to a server creates a privacy concern that isn't worth carrying. Since v4.0.0, everything runs locally.…