What is boundary-first-engineering?

Boundary First Engineering is a written manifesto, not a piece of software, that lays out a set of principles for verifying that code is correct in an era when AI agents are writing much of the implementation. The core argument is that traditional software verification assumed the same person who wrote the code was also checking it, and that assumption breaks down when an AI agent does the implementation and a separate reviewer, human or AI, has to check the work afterward. The manifesto proposes that trust in software should be rebuilt around clearly defined boundaries rather than around the old model of one person understanding an entire codebase. The project is hosted by an organization also behind a separate tool called Suites, which the manifesto describes as one practical implementation of its ideas, though the manifesto itself does not depend on that or any other specific tool or framework. It is meant to stand on its own as a short document, described as roughly five minutes to read. The repository is organized around a handful of files: the manifesto text itself, a lineage document crediting the thinkers and prior work it builds on, a list of people who have formally signed on to the principles, and a contributing guide explaining how to sign, translate, or propose edits. The authors explicitly invite translations into other languages and disagreement in the form of issues or written responses, framing the manifesto's value as coming from being open to challenge rather than treated as fixed doctrine. It is released under the Creative Commons Attribution 4.0 license, which allows anyone to share, adapt, translate, and reuse the manifesto for any purpose, including commercial use, as long as appropriate credit is given to the original authors.