Three trillion dollars move through COBOL every day. Not because nobody replaced it. Because it was designed for something nobody else figured out: encoding business intent so precisely that the machine understands it directly.
When AI models learned to program, they learned on the most semantically dense language ever written. COBOL describes what the business needs — not how the machine works. Every PERFORM, every MOVE, every COMPUTE is a business decision encoded as a statement. AI reads intent. COBOL writes it.
For sixty years, programmers lived between the business and the machine — translating intent into syntax. The business said what it needed. The programmer figured out how to say it in a language the machine understood. AI removes the translator. The business speaks. The machine listens. The code is the requirement.
A COBOL-inspired publishing language for the web. Declare the page — identity, data, structure. The compiler produces a self-contained HTML file. No framework. No intermediary. The source is the requirement. The output is the artifact.
RECALL is not a nostalgia project. It is what COBOL would have built for the web — a publishing language where the source is the intent and the artifact remembers its origin.
COBOL systems from the 1960s still run payroll because they encoded business truth so precisely that replacement means rediscovering that truth from scratch. RECALL asks the same question for the web: what does a page look like if it is designed to last, to be read, and to remember where it came from?