| … | … | @@ -2853,6 +2853,145 @@ The generalization seems to be that stupid people are conservative; |
| 2853 | 2853 | they just vote. |
| 2854 | 2854 | \end_layout |
| 2855 | 2855 | |
| 2856 | \begin_layout Subsection* |
| 2857 | 15 December 2025 |
| 2858 | \end_layout |
| 2859 | |
| 2860 | \begin_layout Standard |
| 2861 | Git commit 2adb566dc7441d95a30171b1c1698a00a8e66a3f message for guix-atomese: |
| 2862 | |
| 2863 | \end_layout |
| 2864 | |
| 2865 | \begin_layout Standard |
| 2866 | Set the GUILE_SITE_DIR for install location. |
| 2867 | \end_layout |
| 2868 | |
| 2869 | \begin_layout Standard |
| 2870 | This is another interesting lesson about Claude, |
| 2871 | actually. |
| 2872 | I had to fight with Claude over this for almost an hour, |
| 2873 | while it proposed all sorts of wild and insane hacks and work–arounds that were obviously flawed. |
| 2874 | (guile doesn't know where guile is installed on a guix system... |
| 2875 | really dude? |
| 2876 | You're gonna go with that explanation? |
| 2877 | And "we can automate this by doing it manually with this automation tool that manually sets the automatic location, |
| 2878 | manually." Uh, |
| 2879 | yeah, |
| 2880 | sure, |
| 2881 | dude.) |
| 2882 | \end_layout |
| 2883 | |
| 2884 | \begin_layout Standard |
| 2885 | Towards the end, |
| 2886 | I was able to force Claude to actually go out on the web and actually RTFM, |
| 2887 | and then if found this very simple solution. |
| 2888 | (which is what is being committed in this commit.) |
| 2889 | \end_layout |
| 2890 | |
| 2891 | \begin_layout Standard |
| 2892 | The experience of using Claude for guix has been extremely frustrating: |
| 2893 | Either guix is more complex than what Claude can easily understand, |
| 2894 | or there is only the thinnest of training material about guix in the Claude training set. |
| 2895 | It seems to know something, |
| 2896 | and is always very eager to offer confident but wrong answers, |
| 2897 | which then have to be fixed by long turn–around time try–fail–fix cycles. |
| 2898 | It's basically shooting in the dark, |
| 2899 | with correct syntax. |
| 2900 | \end_layout |
| 2901 | |
| 2902 | \begin_layout Standard |
| 2903 | This is kind of what I do, |
| 2904 | when I am dealing with a foreign, |
| 2905 | new system that I am not familiar with: |
| 2906 | lots of trial–and–error, |
| 2907 | lots of mistakes and confusion. |
| 2908 | Random efforts with magic incantations to see if, |
| 2909 | this time, |
| 2910 | it will work. |
| 2911 | So it is interesting to see that Claude is doing this with guix: |
| 2912 | a very experimental approach, |
| 2913 | "maybe this will work". |
| 2914 | \end_layout |
| 2915 | |
| 2916 | \begin_layout Standard |
| 2917 | Its very different from Claude's handling of c++ and python, |
| 2918 | where it operates at the expert level. |
| 2919 | Although it is possible that the trial–and–error in those languages is hidden from me: |
| 2920 | when I watch stuff scroll by, |
| 2921 | I see it making vast numbers of obvious errors. |
| 2922 | There, |
| 2923 | I am shielded from the confusion: |
| 2924 | it just scrolls by on a fast iterative cycle. |
| 2925 | Here, |
| 2926 | I am not: |
| 2927 | I get to play the doofus in the middle. |
| 2928 | \end_layout |
| 2929 | |
| 2930 | \begin_layout Standard |
| 2931 | I don't understand the trade–off between memory and ability. |
| 2932 | This is like the early chess machines, |
| 2933 | which simply brute–forced everything. |
| 2934 | Humans take the opposite algorithmic approach: |
| 2935 | remember things, |
| 2936 | because we are unable to brute force (as generalists, |
| 2937 | we have no algo that can brute force things in general.) So, |
| 2938 | I watch Claude an I'm starting to see that it is a brute–force machine: |
| 2939 | it just tries stuff, |
| 2940 | till it gets it to work, |
| 2941 | and usually, |
| 2942 | the results of that search end up being pretty minimal, |
| 2943 | compact and correct. |
| 2944 | Sometimes untidy, |
| 2945 | but I can get it to tidy up by pushing just a little bit. |
| 2946 | Here, |
| 2947 | for guix, |
| 2948 | I had to push a lot, |
| 2949 | and I had to push really really hard. |
| 2950 | \end_layout |
| 2951 | |
| 2952 | \begin_layout Standard |
| 2953 | So ... |
| 2954 | memory ... |
| 2955 | a book of recipes that work, |
| 2956 | to solve some given situational problem, |
| 2957 | coupled to blind–search algos, |
| 2958 | when they don't. |
| 2959 | \end_layout |
| 2960 | |
| 2961 | \begin_layout Standard |
| 2962 | Hmm. |
| 2963 | OK. |
| 2964 | \end_layout |
| 2965 | |
| 2966 | \begin_layout Standard |
| 2967 | — |
| 2968 | — |
| 2969 | — |
| 2970 | |
| 2971 | \end_layout |
| 2972 | |
| 2973 | \begin_layout Standard |
| 2974 | Yes, |
| 2975 | I wrote the above into a commit. |
| 2976 | I spent the last 24 hours doing something that I expected to take an hour. |
| 2977 | Claude has been consistently throwing me under the bus, |
| 2978 | with this guix project, |
| 2979 | which is not what I have come to expect. |
| 2980 | \end_layout |
| 2981 | |
| 2982 | \begin_layout Standard |
| 2983 | On related news: |
| 2984 | I don't know if this was an accident or what: |
| 2985 | it was unable to write out the Gnu COPYING file. |
| 2986 | It just wouldn't do it, |
| 2987 | and notice that it didn't do it, |
| 2988 | and then try to do it, |
| 2989 | and not do it again. |
| 2990 | I wonder if this is some safety–engineered prompt in Claude, |
| 2991 | that prevents it from accidentally sticking the Gnu License on things where it might not belong, |
| 2992 | and where the user might not notice (because COPYING is such an innocuous file name...) Interesting. |
| 2993 | \end_layout |
| 2994 | |
| 2856 | 2995 | \begin_layout Section* |
| 2857 | 2996 | The End |
| 2858 | 2997 | \end_layout |