Yeah but you didn't have internet to waste your time back then. I fucking hate internet at times. It has conditioned me to read everything in little digestible tidbits. Before I spent much more time reading actual books, staying concentrated on reading one topic for you know at least 1 hour at a time.

Sure we did. I was very active in several Usenet newsgroups from 1987 on.

Of course, at that time, few companies were connected 24x7. My employer, for example, dialed in for mail and Usenet traffic exchanges a handful of times a day.

The debates over the first ANSI C spec in comp.lang.c were most entertaining.

The two are not totally incompatible. If you can decouple segments of huge projects, you can mock input and get quick feedback on the specific segment you are developing. Insanely long builds still exist today, I know of and have worked on software that ranges from 1 hr to 1 day to build quite recently...you just have to find ways to avoid that huge cost :p

You still get that in machine learning projects - you make some change, launch the re-training of the models and see the result next day, or in an hour, or in a week; and only then you know if the change was actually beneficial.

You lucky bastard!

I remember the joy when we got Cadence BuildGates to start running parallel synthesis and optimization runs. This allowed us to go from one build per week to two. If we were lucky!

Some of us have an idea, if only because a full build and integration test cycle for clusters of virtual machines can take hours depending on how the pipeline is organized!

Pffft. They have no idea what it's like to work on really big software projects.