• I’m working through ~14 old hard drives I’ve accumulated over the years from various computer upgrades and decommissioned home NASes.

    Each one is getting dumped, merged, and pruned until I have exactly one current, unified file tree representing all of my data. No more “old-w540-ubuntu-2017/” folders where I’m not sure if that’s the only copy of the data or not.

    I’ve misplaced some data over the years, and my laptop drive is always full. I hope to address both problems.


    score 1 · canonical

      • I’ve got a 4-bay QNAP configured in passthrough mode. I got once drive cleared of and now it’s holding the dumps. btrfs RAID 1, with deduplication because there’s a lot of overlap between the dumps. As I clear & verify more drives I’ll chuck them into the array.


        score 1 · canonical

    • No more “old-w540-ubuntu-2017/” folders

      I’m in this picture and don’t like it :)

      Is there any specific tooling you’re using for this endeavour? There’s a big tote of drives sitting on the floor here next to me that deserve the same treatment…


      score 1 · canonical

      • lobsters__spiffytech@crustaceanson.theshelf.topM
        link
        fedilink
        arrow-up
        1
        ·
        edit-2
        11 days ago

        btrfs deduplication helps cut down on intermediate storage needs while I get all the file trees in one place so I can compare them.

        Comparison is tricky since for trees evolve over time. I need to detect when two folders at arbitrary places in the file tree have highly similar content. I’ll roll up arrays of rapidhash values and throw each hierarchy level at Levenstein and see how far that gets me.

        But I’m also going to try directly accessing the btrfs block identity data instead of doing a manual scan.


        score 2 · canonical

        • Dope! Yeah that’s going to be the big thing for me: drives from 5 years ago are going to have some content that isn’t on my current drives and some content that is a git repo that I do have on my current drives but not with the 5-year-old checkout. At least for those kinds of things I should be able to check the remotes and see if they match


          score 1 · canonical

  • Hopefully finishing up two circuit board layouts that have been lingering since… January or so. One is for Christmas lights and Christmas is coming faster than I’d like.

    Other big thing is to start getting ready for winter. One last mow and yard cleanup. Put gas in the snowblower and see if the carb needs to get cleaned or not. Put up markers again to indicate where the baby trees are so I don’t take them in January.


    score 1 · canonical

  • After two non-stop days of rain, I’ll be enjoying walking through the city with my wife. Trying to hit a few dives, a bookshop or two, and get some food somewhere.

    Also cleaning and practicing system design


    score 1 · canonical

  • Thinking about a job offer. I enrolled for the master’s degree you need here to be a secondary school here, but an ex-colleague prodded me to apply for his company. They made a decent offer and in theory they would accommodate my studies.

    However, my mind was already set on focusing on the master. I was very disenchanted from my previous gig and the master’s is a plan B to have job options outside coding.

    My ex-colleague’s company is apparently friendly to people who won’t use LLMs for now, and I’m told that only one of the members of the team I’d be going in relies heavily on LLMs, so it’s not a bad option and well, getting paid while I study wouldn’t be so bad.

    I’m considering writing them why I’m taking the master’s. If they still hire me after that (doubtful) then I would be more comfortable joining. (I know it’s not really a recommended strategy to be this kind of transparent to prospective employers, but… I was not even considering a job, so if I lose it, no big deal- and if they hire me anyway, that’s likely a great sign about working there.)

    Other than that, probably I should code a bit. I’m mulling:

    • Implement kind of a real-time relational database on top of LMDB. I think I have a valid design (here and I think it’s an area worth exploring.

    • Continue work on a small Nagios replacement I started drafting. Nagios is much more awesome than most people think, but I’ve become obsessed with a dumb issue with it: to configure a service to take more time to send a notification for warnings than for critical status requires absurdly verbose configuration.

    • I also had an idea that I thought I could implement a kind of version using rclone serve s3. The idea is to provide libraries so that applications can have configurable remote storage; like you use the library in your app and the first time you need to store data, the user gets a dialog so that they can choose some remote storage (Google Drive, Nextcloud, etc.) and grant the application access to a single folder inside that. I feel this would allow application writers to add “sync” functionality to their apps without having to run any backend service. Another idea worth exploring. But this requires so much more effort than the other two that I don’t really want to do it.


    score 1 · canonical

  • I’m working on doing my taxes, because I don’t want to leave it until last minute, and also I want to pay taxes as I go as I don’t want to get stung by interest fees again >:(

    I’m also reading a very interesting book: “The unwomanly face of war”, which I’ve borrowed from the library.


    score 1 · canonical

  • I’ve been working on Lunacy, my interpreter + JIT compiler for Lua, again recently. It’s now (very) fast, but I need to do a pass over all the Claude-slop quality code and rewrite the bad parts and all the comments. Then once I have all the new features cleaned up I can actually merge it into main and write a new blog post: I want to brag a bit!


    score 1 · canonical

      • lobsters__chc4@crustaceanson.theshelf.topM
        link
        fedilink
        arrow-up
        1
        ·
        edit-2
        10 days ago

        I wrote an initial blog post about it a few months ago at https://redvice.org/2026/lunacy/ - I think at that point it was like 95% written by hand, because LLMs were simply too stupid to actually work on it effectively and would immediately break the JIT and then claim it was broken the entire time and never worked. Now they’re a bit better, and I’m able to have Claude Opus 5.5 do most of the programming work and refactoring which has been convenient. It definitely still makes commits doing things the wrong way, and most of its “suggestions” are things that actually completely defeat the JIT design or several important invariants so it still can’t be trusted to make any sort of decision, but for a fun side-project that I’m doing mostly to mess around with neat JIT techniques it’s been nice to have it do the mechanical work.

        Do you have any public material about your JIT? I’d love to check it out!


        score 1 · canonical