Not a single invention, but a coherent progression — each step building on the last, from a new way of making chips to a system that continuously improves them.
The earliest public identity of the semiconductor platform that became the foundation for the wider technology — a different route to building logic, not just shrinking it.
A series of early device and computing architectures demonstrated it was possible to move beyond conventional CMOS assumptions — and to reduce overall transistor count through alternative logic, rather than relying on geometric scaling alone. Work expanded from individual devices into computing systems.
Manufacturing methods were developed to turn device concepts into repeatable fabrication processes — shifting the work from theory toward something that could actually be built.
Logic functions implemented directly within atomic-scale tunnel interfaces, plus methods to increase switching speed. The key idea emerged: speed improvements could compound logic-density improvements.
A new transistor architecture made tunnel-based logic practical through very fast, low-voltage switching — and additional device architectures expanded the available design space. Improved isolation structures made the devices more robust.
Tunnel logic was integrated directly with the new transistor architecture — combining the transistor and tunnel-logic innovations into a more complete whole.
Three-dimensional extensions increased integration density and opened up future scaling options — building upward, not just smaller.
The enabling breakthrough: compressing semiconductor development cycles from months to days. Same-day manufacturing is what makes a continuous self-learning loop practical.
A new memory approach that scales through depth — additional layers and fabrication cycles — rather than through ever-finer geometry. Conventional memory scales through geometry, months and masks; this scales through depth, cycles and repetition.
A hardware architecture designed to improve through repeated manufacturing and operational feedback — treating hardware not as fixed, but as part of a managed improvement cycle. Intelligence improves the machine it runs on, then runs on the improved machine.
AI, hardware design, manufacturing and testing connected into a single controlled improvement system — using measured results from one cycle to help determine the next.
A new foundation, then logic that needed fewer transistors, then the devices to make it practical, then building in 3D, then manufacturing in a day — and finally a loop that uses each result to improve the next.
The progression matters more than any single step: the strategic value is a system that continuously improves designs, not one particular design. The full patent record, with dates and sources, is on the IP & terminology page.