The patent system is dysfunctional in large part because it consists of too much text for humans to reliably comprehend. Recent developments in artificial intelligence – specifically large language models – offer the potential to automate significant amounts of intellectual labor inside the patent system, thereby making it more legible and functional to its participants. What makes this problem particularly difficult is the precision of patent language. To the extent that patent language is among the most precise English in the world, out of the box large language models are not able to (and for the foreseeable future will not be able to) automate patent inference at the level of accuracy required to be meaningful. Notwithstanding, were one to design a bespoke system that properly structured and orchestrated the thousands of atomic thought processes involved in patent inference, one could conceivably build a system immune to hallucination and logical inconsistency. We have spent the past three years building this system whose first implementation is a high-end prior art search service that is superior in affordability, reliability, and speed to any other search product on the market. Our intention with this flagship product is to build a durable brand that facilitates both (a) the procurement of government contracts to automate the patent examination process, and (b) the engineering and distribution of other products and services – harvesting, prosecution, freedom to operate, monitoring, litigation, licensing, etc. – each of whose automation is substantially similar to the automation of prior art search. In doing so, we intend to reinvent the patent system.