A sketch that might inspire brainstorming. I perceive that we're "doing what we've always done" instead of stepping back and inventing new stuff, like they did in 1960. The "we've always done it this way" mentality results in epicycles piled on top of epicycles.

For some, it may be "doing what we've always done". But I suspect that for many more people, it's "this is what commodity tech is today, let's build interesting applications on top of it". And the more people rely on your tech as infrastructure, the harder it becomes to replace with something radially new.

We've invented a whole lot of stuff since 1960 not just LLM's, for example strings (GCed), hash tables, queue classes, overlapping windows, closures, etc. Many will continue to look at this stuff through a 1960s lens, whereas a few might use a 2025 lens (I hope). No radically new tech, just a different set of mental biases.

For me, the baseline is not 1960, but 2000, when personal computers became a commodity, followed closely by the arrival of the Internet. We had strings and hash tables by then, so they became part of the commodity toolkit. From then on, technodiversity went down. We had lots of processor architectures in the 1990s, but only x86 and ARM survived the transition to commodity tech. Among the architectures that disappeared, the transputer is probably the attempt closest to your ideas.

Yeah, I was quite interested in the transputer and OCCAM. I perceive there being a hangover from the 1960s that we haven't shook ... so I guess that I think that commoditization happened much earlier than 2000 (I blame "C")... https://programmingsimplicity.substack.com/p/weve-always-done-it-this-way?r=1egdky

The X axis is time. What's the Y axis?

Technological inertia. Which I think is roughly proportional to the number of non-tech users.