Shaping the Loops, Closing the Loops

(revyl.com)

30 points | by based0x10 3 days ago

5 comments

  • bill10 17 hours ago
    Good loops increase efficiency, bad loops produce more garbage
  • bombela 19 hours ago
    In summary: Removing friction for an increase in efficiency.
  • anamhira 3 days ago
    Huge move to working at higher level of abstractions to have the ability to have more leverage.
  • cootsnuck 18 hours ago
    I keep seeing people rediscovering tried and true software engineering principles and saying AI makes them "matter more"... And I'm a bit skeptical of that framing.

    Things like platform engineering, DevOps, code quality, linting, CI/CD, QA, etc. were just as foundational before coding agents as they are now (regardless of never being called "pattern engineering"). I'd argue what's changed is less their importance than the consequences of neglecting them.

    I think a lot of this is how we as an industry are coping with the realization that the narrative we've been sold over the past few years, that AI was going to render software engineering a "solved problem", isn't actually true.

    So instead of saying "Hm, it looks like being able to arbitrarily generate massive amounts of plausible code has exposed just how much of software engineering was never about writing code in the first place"... We say "Wow, these amazing AI coding agents are so powerful we're having to formalize new levels of abstraction for engaging with software engineering!"

    Idk, maybe I'm being a yum yucker. People can call stuff whatever they want. Just irks me when there's plenty of prior art, knowledge, instruction, etc. that has run into these problems years and years ago that may be more useful to center than coining new shibboleths.

    • kyralis 18 hours ago
      I think the difference is in effective bottlenecks. Previously, the ability actually produce code was a bottleneck just as much as the ability to architect and engineer that code, or to make effective product decisions was. Now, the code production bottleneck has been largely reduced - but those other bottlenecks have not been, bringing them to significantly larger prominence.

      The reason this matters for baseline processes and patterns is that when the system is not set up for AI to trivially copy good practices, it won't, and that will cost the architect and engineering bottlenecks even more just at a time when they are becoming the most obvious limit on velocity.

      So, yes, they always mattered for lots of reasons. But the relative impact on velocity, which is at the forefront of everyone's mind right now, has changed.

      • cootsnuck 17 hours ago
        I've had a different experience in my career. The ability to produce code was seldom the bottleneck.
        • jghn 4 hours ago
          Same. When I hear someone say that rote typing of code was their bottleneck, what I hear is that they weren't involved enough in design decisions, did not think deeply about how they were approaching the problem, nor did they spend enough time self-reflecting on what they just wrote. Over the years I worked with plenty of developers who'd just take a ticket without thought, bang it out as quickly as possible, and just throw it over the wall without a second thought. These are the types I assume are saying this stuff.
        • tonyedgecombe 15 hours ago
          It seemed to be a problem for a lot of programmers. Maybe not the people here but the average enterprise developer wasn't very productive at all.
  • nateroling 17 hours ago
    I’m reminded of Jack Sparrow sailing into harbor in Pirates of the Caribbean, proudly standing atop his rapidly sinking mast.

    Here we are, the pattern engineers! High atop the ladder of abstractions, clinging to relevance while the AI ocean continues to rise toward us.