3 ms·
> Why isn't the next layer of abstraction simply what is being termed as "No Code" today? Cynically? Because people were also selling "no code" ten years ago,
by Bartweiss 7y ago
> Why isn't the next layer of abstraction simply what is being termed as "No Code" today?
Cynically? Because people were also selling "no code" ten years ago, or even further back.
I know that sort of inductive argument doesn't actually work. We went through a dozen nonsense models of infections, but germ theory was actually correct in the end. AI has seen frauds and baseless hype since the mechanical Turk, but there still came a point when computers won at chess and Jeopardy. But it's a pretty good warning sign.
The gradual progression of abstractions still caps out somewhere around Python and Rails. We have drag-and-drop website makers, and visual tools for "plumbing" ML flows backed by pandas, or engineering calculations backed by Matlab. They mostly work, and people who know the underlying tools sometimes use them for speed while mixing in actual code as needed.
Meanwhile, promises of codeless business apps don't seem rooted in any of those advances, and continue on from their dysfunctional predecessors. And people actually making business apps in a hurry pull from NPM or PyPI without ever touching "no code" tools. It seems like we might get much closer to that layer within a decade or so, but I don't think we're there yet.
- coffee 7y ago> Cynically? No, not cynically :-) > Because people were also selling "no code" ten years ago, or even further back. Not really sure it matters when it starts. Sometimes markets aren't ready and products are too early. What matters is are these solutions providing value to people today, and is it enough value for them to buy/spend time on/etc... Seems there is enough value today, with a mature enough market, based on the results these platforms are experiencing. > I know that sort of inductive argument doesn't actually work. Then why waste time making it in the first place? :-)