The Specificity Gradient (Rough Cut)

9 min

FULL-LENGTH CONFERENCE TALK VERSION: https://youtu.be/iiRdzSVKUeI . SEE ALSO https://doriantaylor.com/the-specificity-gradient . Here I suggest there is a set of "shearing layers" (aka "pace layers", à la Stewart Brand) that arise in the process of developing software, and advocate that the artifacts that lead up to the code are profoundly undervalued relative to the code itself. What if we kept these things around, and could traverse the specificity gradient that arises naturally in a software project, which ultimately reduces to figuring out how to take a desired outcome to an ultra-precise method of achieving it? There is an entire documentary about shearing layers: https://youtu.be/HTSbtM12IZw I should also remark that the original shearing layers framework was Frank Duffy (https://en.wikipedia.org/wiki/Frank_Duffy_(architect) ), not Stewart Brand. 00:00 - desired outcome → specific method 01:12 - business goals → user goals 01:29 - user goals → user tasks 02:03 - user tasks → system tasks 02:31 - system tasks → system behaviours 02:45 - system behaviours → code 02:51 - we forget about the value of everything but code 03:16 - shearing layers/pace layers as precedent 03:44 - the original shearing layers 04:57 - the generalized pace layers 05:49 - each category lives in its own temporal domain 06:13 - we have an analogy in the specificity gradient 07:02 - code is actually disposable 07:42 - how do i see all the way from business goals to code and back? 08:21 - if something gets obsolete, you hack it off and make a new one 08:27 - imagine this structure accumulating over decades