The Abstraction Level Changes With Ai
We have always moved up the abstraction level at which we build software, as we made better tools that helped us build faster and better software at scale.
We have always moved up the abstraction level at which we build software, as we made better tools that helped us build faster and better software at scale.
Twenty years ago, if you wanted to create an online shop, you would have probably coded the page yourself entirely, found a server and hosted it.
Then came the cloud to simplify hosting, and then came tools like Shopify which streamlined ecommerce for the web, giving small merchants the same software that Amazon has. Well, not at the same scale, but a version of it.
The reason companies like Shopify make profits is because they take the best practices of building a website at merchant-scale and apply it to merchants worldwide. And since they do it at scale, they gain the benefits of centralization.
But as the tools we have improve, the abstraction level we operate at also goes higher. Instead of writing your own code and hosting it, you just put in your credit card, click a couple of times and your shop is online.
Now I've had this theory. Why are we reinventing things when the best practices are already known and documented by people smarter than you or me?
Take the Amazon example.
Did Walmart have to buy Jet.com and shut it down to build its best ecommerce version of itself?
What if the best practices of Amazon, a worldwide scalable website that handles hundreds of millions of users with no lag and no downtime, is already written down somewhere in the collective genius of all the developers who worked there over the last twenty years? What if that knowledge could be curated, templatized, and made readable by AI?
Would that mean the next Walmart could use that template and spin up an AI version of Amazon in a fraction of the time? Yes.
There are still hurdles to this outcome though.
The collective knowledge of building Amazon is not in one single place. Curating the right ideas will require human architects who are well-versed with the complexities of cloud computing. But if you curate that knowledge and theoretically open source it, AI should be able to take it and run with it.