Pascal: A new medium for enterprise software

For the last two decades, we’ve been building static software for organizations regardless of each organization’s mind architecture and operations. Slack is the same product regardless of whether you’re a ten-person organization or an organization with tens of thousands of people. It’s the same whether you’re an investment firm or a tech company. That wasn’t the original premise of software. The early days of software were filled with tens of projects attempting to address the static nature of software: the semantic web, Microsoft’s Business Objects and COM architecture, CORBA, enterprise knowledge management systems like HCL Notes, and early AI-driven expert systems like XCON and CLIPS. But all these systems failed because software couldn’t understand language. This has changed, and the way we think about software needs to change. AI platforms today, however, are merely the old software with smarter search. We now have the technology to automate entire departments, but we don’t have the infrastructure layer for it. Companies like Atoms articulate the idea that the next frontier is not bits or atoms in isolation but their convergence: treating the physical world as a computable system in which sensing, prediction, and control form a continuous loop. There are also two worlds inside the world of bits itself: systems of record that reflect an organization and the knowledge, intent, and coordination that drive it. We store the latter but rarely treat it as computable because we historically lacked intelligence as a computational primitive. The current limitations stem purely from the medium. LLMs need access to the geometric and operational substrate of the real world or the world of judgement. Intelligence needs to be paired with structured, ontological, and operational systems so LLMs can both understand and intervene in the environments they reason about. Pascal is structurally built on the thesis that the generality of LLMs and the problem of context demand centralization and an architecture equally general to the technology. I believe that the geometric and linguistic infrastructure should exist on a single platform to do for every industry what LLMs did to coding.