Two weeks compressed the quantum clock. Nobody is ready for what… — Cassiopeia — TG.ME

Two weeks compressed the quantum clock. Nobody is ready for what comes after. The qubit count needed to break internet cryptography fell by 40x, then again by 20x, within a fortnight — and that is not the surprising part. Since Peter Shor's 1994 paper, public-key cryptography has had an invisible expiry date stamped on it. RSA, ECC, Diffie-Hellman all rest on problems that a large enough quantum computer factors in polynomial time. For three decades the comfort was simple: the machine did not exist yet, and building it looked very hard. That comfort is dissolving now because the distance between "theory" and "engineering" has started to close. On 31 March Google Quantum AI revised down the qubit cost of breaking ECDLP-256 by 20x — now under 1,200 logical qubits, about 500,000 physical. The same day, ex-Google/Caltech researchers at Oratomic argued Shor's algorithm fits on as few as 10,000 reconfigurable atomic qubits, a further 40x compression. Two independent teams moving the threshold in the same window is not a fluke estimate but a reassessment — and Google explicitly framed its paper as a call for faster post-quantum migration. What made the reassessment credible was Google's Willow chip in late 2024. For the first time, surface-code scaling worked the way theory predicted: adding physical qubits per logical qubit halved the logical error rate, rather than raising it. Before Willow, error correction got *worse* with scale; after Willow, building a fault-tolerant machine became an engineering problem rather than a physics question. The risk frame that makes this consequential is "harvest-now-decrypt-later." Encrypted traffic that adversaries record today becomes readable the moment large quantum computers arrive, and a 2025 Telecom paper formalized HNDL as a *temporal* risk — every month of migration delay is a growing liability, not a steady-state one. NIST's August 2024 FIPS 203/204/205 standards named the destination; Google has publicly set an internal 2029 deadline to get there, six years ahead of the US federal 2035 target. Cloud customers inherit that deadline whether they planned to or not. The commercial side is moving at a different tempo. Xanadu SPAC-listed on Nasdaq on 31 March and opened +15%, following Horizon and IonQ, but Horizon is -18% since debut and Infleqtion is down 30%+ since its February NYSE listing. Capital wants to own whichever substrate will matter in 2030 — superconducting, neutral-atom, photonic, or trapped-ion — and public markets cannot yet price a decade of waiting. Bain projects commercially meaningful demos (100 logical qubits) by 2028-2029, and real applications (1,000-10,000 logical qubits) in the mid-2030s. That is the central tension: the cryptographic threat is arriving on a faster timeline than the commercial payoff. spacefrontiers.org #quantum_computing_era Numbers, sources, and triggers — in comments ↓