Software is complex and made from slime.

It is complex both in the everyday sense, and in the more technical sense used by complexity science. It is slime in that software is continually forced to engage with real world systems very unlike the laws of highschool physics, systems where rules are very local and contingent. William Wimsatt calls this "ontological slime".

This complexity is why agile software development and design are effective, including the usefulness of human feedback and of reducing localised complexity in code. We also have an agile idea, from Peter Naur, that a large part of programming is building and improving a model of the system in the programmer’s head. What does that imply about the worldview of an effective programmer? Because software is a complex system, a great developer needs survival techniques for navigating and building beautiful machines out of causal thickets and ontological slime.

Evolving an organization to use more agile techniques often means re-evaluating the role of dedicated QA teams common in waterfall development processes. In software organizations that develop a large platform with a group of agile development teams, there is still a need to ensure quality at a product and platform level. As one solution to this, this talk introduces the idea of Platform QA, a dedicated team of QA specialists with responsibility to the entire product, and the platform that delivers it. Platform QA is a feature team that works on a backlog of quality risks and has ultimate ownership of shared artifacts such as end to end (BDD) tests and QA environments. This talk discusses differences between waterfall QA, platform QA and embedded QA, examples shared from my own experience, and conditions where this solution may make sense as a transitional or target organizational structure.