Quantum computing is not just moving fast; it is moving increasingly faster. Hardware fidelities, error-correcting codes, compilation techniques, and algorithms each advance on their own timeline, and progress in any one can reset the economics of the others.
Consider resource estimates: the number of physical qubits thought necessary to factor a 2048-bit RSA key fell from roughly 20 million in early 2025 to approximately 20 thousand in late 2026, largely through algorithmic and architectural improvements rather than better hardware. AI-accelerated research is compressing these cycles further. Machine learning systems now decode errors, optimize circuits, calibrate devices, and search design spaces for new codes and algorithms, turning research loops that once took months into ones that take weeks or days. Each improvement feeds the next. Human experts can no longer even keep up with the progress with an average of 70+ new papers arriving in the scientific literature every day.
The result is a field whose frontier moves faster than conventional planning cycles. The cost of making a decision based on old intelligence is substantial: not only must you live with the delay of not making progress in the right direction, but you must also live with the cost of moving in the wrong one.
The implication is straightforward. When technical inputs change, the evaluation must update, and it must update quickly. Intelligence cannot be a document. It must be a system.
