Clojurians Log v2

I started typing about tacit knowledge and it threatened to become a blog post, but https://www.evalapply.org/posts/after-ai/index.html#main... TL;DR: All expertise grows logarithmically, not exponentially > hard-earned specialist tacit knowledge, is entirely absent from any static data corpus, no matter how detailed the corpus is. Taste, judgement, and intuition are not born, they are bred. Through diligent training, deliberate practice, and frequent feedback from violent contact with reality. Frequent violent contact. > > However, Human and Computer expertise grows along logarithmic growth curves. Gains grow quickly in the beginning, for a short while, then slower but still perceptible for a bit more, and then, suddenly, imperceptibly. No amount of additional input produces any meaningful additional output. > > Said another way… Expert systems need expert users.

> The West Forgot How to Make Things I guess this title is more catchy than what the article describes: The west spent 30 years not making bombs and ramping up production was slower than promised. > By the deadline, Europe delivered about half. > When Knowledge Dies, It Stays Dead Per wikipedia: The W76 life-extension project was completed in December 2018, when 800 W76s were upgraded to the W76-1 design.https://en.wikipedia.org/wiki/Fogbank#cite_note-9 It is unclear whether the new W76-2https://en.wikipedia.org/wiki/Fogbank#cite_note-10 uses Fogbank. It's unclear whether the knowledge was actually lost. It's also not clear whether Fogbank even continues to be necessary. The article makes the obvious point that building and coding require continued practice to be effective, but then it hand waves when it comes to the interesting bit. Will the techniques that developers are no longer practicing be necessary in the future? I'm not sure, but the article doesn't make a good case either way. I think much of coding today is completely incidental. If junior developers aren't spending time making their own bashrc and bash_profile and knowing what the difference is, that doesn't seem like a problem. The article also doesn't have much to say about where all the bomb making expertise went. Maybe it's possible that these folks were applying their skills somewhere else. I can at least imagine that it's possible for folks to be more productive doing something else besides wrangling kubernetes configurations.

I don't disagree with your take. Your bashrc example is a strawman though. Understanding that may be incidental, but understanding what a shell even is may not be. The article is quite sensationalist and obviously LLM-assisted (hopefully, as an aid for an ESL author). But it conveys what I think about the dangers of overrelying on alien technology, with examples that are understandable and, unfortunately, very relevant to me too. I have a junior colleague who asks claude code to find all references to a variable in the project. He doesn't know about grep or IDE project-wide search. Is this what we as a society want to spend drinking water on?

Your bashrc example is a strawman though.I could probably spend hours listing examples of arcana that is completely incidental to producing software and never needed to exist in the first place.

> I have a junior colleague who asks claude code to find all references to a variable in the project. He doesn't know about grep or IDE project-wide search. Is this what we as a society want to spend drinking water on? I'm not saying you couldn't make the case that our current use of AI will lead to a knowledge gap in the future. I'm saying that the article doesn't do that.

And you would be right, yet it seems like chatbots will never get rid of that arcana, they will continue to stack upon it ad infinitum because they lack the anergy and distaste for it. It is akin to how we coded ourselves out of x86 idiosyncrasies with better high-level languages and compilers. It doesn't mean that x86 shouldn't burn.

That's just for PC. I believe phones are mostly arm.

As someone getting older, I'm sympathetic to the idea that the next generation, with more and better tools/resources, will somehow be less useful, but I'm skeptical.

Yes, I'm trying to make a point that the transition to ARM is happening because people involved still care – about performance, about complexity, about DX. Chatbotting a problem lacks those urges (at least, for now). Who cares if React is a bad tool for the job if an LLM will make it fit into any Procrustean bed? > As someone getting older, I'm sympathetic to the idea that the next generation, with more and better tools/resources, will somehow be less useful, but I'm skeptical. Ironically, one of my fears is that the next generation will be stuck using the inferior, clunky tools and resources of your and mine generation instead of creating better ones.

the inferior, clunky tools and resources of your and mine generation instead of creating better ones.emacs and the vt100 are from the previous generation. I'm not that old! 😛

Yea, I agree that it's an important discussion and we should be thinking about the consequences of how we use (or don't use) AI now and in the future.

Technical debt has traditionally been something that was accrued over many years, in steps given by the initial developers leaving the project or the company. These days, with the help of AI it's possible to get the same technical debt from the first minute 🙂

I finally finished reading this series, https://aphyr.com/posts/411-the-future-of-everything-is-lies-i-guess, which was very pessimistic.