I built a Bitcoin signing device where the private key is physical

The core problem I kept running into with hardware wallets: the private key exists as a persistent digital object inside a chip. It’s protected — but it exists. That’s the vulnerability. Protection fails; existence is structural. I built a different model. The private key is encoded as a geometric hole pattern in a titanium plate. A signing terminal reads the plate optically, derives the key transiently in volatile memory, signs the transaction, and discards everything. The plate doesn’t change. Nothing is stored. A few implementation details that might interest people here: The signing terminal uses two physically isolated MCUs — one handles communications, one handles signing. They are not software-separated. The communications MCU is physically disconnected during key derivation. This isn’t air-gapped in the usual sense; the isolation is enforced at the hardware layer, not the software layer. The key derivation follows BIP-39/BIP-32. The plate encodes the seed phrase as a geometric representation. The terminal reads the geometry, derives the entropy, derives the key, signs, clears volatile memory. The plate never changes state and has no electronic components. The threat model this addresses: persistent digital key at rest. Hardware wallet attacks — glitching, side-channel, physical decapping — all require the key to exist somewhere in the device. If it doesn’t exist, those attack surfaces don’t exist. The attack window is constrained to a single observable physical interaction, not an indefinite future. Patent application published: US 2026/0039478. Pre-orders open at frozensecurity.com. Happy to go deep on the architecture, the encoding scheme, or the threat model.