• 0 Posts
  • 233 Comments
Joined 6 months ago
cake
Cake day: June 5th, 2024

help-circle





  • Yeah, one of my projects right now has been delivering huge value with very few staff-hours being expended in coding. That’s because I (senior architect) and a couple software engineers researched the shit out of it before we started, and found a way to adapt free, existing, running code with minimal effort. I’ve seen two previous attempts to do this job fail expensively and catastrophically. So far, we’ve spent 15% of what either predecessor project cost, and we’ve already got operational code deployed and a solid proof of concept for the rest. That’s because of months of hard thinking and experimentation by my engineers and me. And yeah, that’s right, it meant doing some Big Design Up Front, and fuck you to every agile fanboi who thinks you can accomplish a highly complex integration project without doing that. We’ve already had a couple of those knobheads lose their jobs for failing at previous attempts, then opposing my approach. I’m hiring more real engineers with the freed-up headcount.

    Some of this work is irreducibly hard and anyone who thinks they can factorize it into a bunch of parallelized trivial processing doesn’t know the problem space. Snitchware and truncheonware are not going to change that.





  • Frankly anyone telling you they can measure the value of a line of code without any background knowledge is selling BS.

    the previous system of managers not so secretly counting total commits and lines added was comically stupid

    That has been known not to work since the 1970s. There’s probably something in The Mythical Man-Month ridiculing lines of code as a performance metric.

    Some of the most productive work I ever did involved ripping out 80k lines of executable code and replacing it with 1500.

    But I welcome this new BS system

    I don’t. Fuck snitchware in all its forms.








  • Whether a platform is federated or not is an abstract and irrelevant question to most users. It’s like telling a typical end user that their hardware architecture is big-endian or little-endian. In terms of their usage, it makes no difference.

    federated platforms are a major failure, as picking instances and creating accounts is a hassle rather than a convenience

    Some UI improvements and simple affordances could make a bit difference to those barriers.