Programmable glue

Coding is undergoing a metaphorical restructuring.

Until recently, code was best understood as stacked building blocks. Coding, therefore, meant one, or a combination of, three things: 1) creating a new block, (2) packaging code into a block, (3) stacking two or more coding blocks.

Coding blocks have been themselves varied in size and function. A block could be a function, a library, an SDK, an API or a stand-alone product or service. Even an entire company could be a building block, like Crowdflower.

Naturally, we talk about software product as built with “tech stacks”. What else can be stacked if not blocks?

Blocks were commoditized quickly, giving rise to bigger and more complex blocks. This, in turn, gave rise to the need to “orchestrate” and “architect” these building blocks. And with this came more and more complex structures made up of more stacked blocks.

AI amplifies a trend that started with so-called server less functions. Logic that runs on demand, so that blocks don’t need to be ensconced in the stack but in turn

Coding will be more monstrous, in the Venkatesh-ian (Rao) sense because we can now program glue. Glue plays well with existing blocks, but it can also adhere itself to new paradigms, as well as new metaphorical understandings of what code, and intelligence, can be. Glue can hold different blocks. Programmable glue can be reconfigured to hold some blocks together at some point, and another set of blocks together at a later point.

Programming glue means we spend less time creating new blocks, or stacks of existing ones. It means we mortar blocks on top of blocks, but we also experiment with new ways of stitching them together, so that stacks becoming only one possibility.