fin1.019.sitrep 20251107 https://codeberg.org/Digit/fin [Edit, not so much a fin sit-rep as much a me sit-rep (not just as lead fin developer).] well, writing this from fin1.018.fishsticks.fish. it's fine enough. not too flickery. not like how the progress has been going on fin1.019.00*.fish. it's flickery. has the pseudocursor visably scroll down the leftmost collumn, the whole way and fin1.018.fishsticks.fish has the proof of concept working undo/redo. it works. ... i think that's what reintroduced that little extra flicker. but at least it's not like the 019 branch(?) in my dev dir. my scruffy scruffy dev dir. tho, with my git-fu, it's easier to keep track of. scary that i basically let the llm, (using opencode (and grok, iirc)), change the files in my git dir, for fin1.018.fishsticks.fish ... and i confess, i did not scrutinise the code as closely as i have prior. before that move, that first opencode move, i scrutinised. i not only scrutinised, i persisted, insisting it done right. done the way i imagined. and... well, it was only a few back and forths, not multiple bouts of >30 back n forth, so when it seemed to have something working, i basically went with it. i saved my sanity. but at what cost? lets hope here after, (now with opencode working on my files in my dev dir, not my git dir), i am back to scrutinising the code. ... and not atrophying my skills. the point with this [using llm to program [*1]] is to improve my skills. [*1: not fin itself. the point of fin itself, is fin itself. fish extensible text editor.] ... the extensibility first pokes, didnt go well... need to talk it through with other llm, not just the coder have had too many thoughts about what to tackle next. several queries to grok via the opencode interface[*2] [*2: frankly, none better for typing in than fin is at this stage. weird huh? for nvim.] more likely shall make better progress with qwen, mistral, and claude sonnet, on those discussions of what to prioritise to streamline. especially qwen, methinks. qwen's happy swallowing a boatload of info and files to fathom. yet to get local MCP tho... scary as that is to say, having grown up with tron. there, MCP isnt Model Context Protocol. it's Master Control Program. ... LLMs and MCPs tellin me what to think. ... ... it's like trying to tell Quentin Tarrantino how to do things, on set. Just no. We're doing it right. No we're not going to make a bunch of new bugs. Yes we're going to do it right. Though, I am flexible enough to concede to better ideas, since this all started on a whim, and i dont really know the best ways to do things. low programming aptitude. 20 years programming (or nearer 40 if you include a couple lines of spectrum basic in the 80s), and still suck at it. 20 years even with haskell! ... maybe that's why i suck. ;D but, contrary to popular opinion/experience, i've been learning way more fish, since the fin idea struck me, and i got an llm to do as i said. learned not just fish, but more about core utils, and even got a glimpse at awk. i never glimpse at awk. awk scary. awk made less scary now. good llm. good llm user, not just swallowing whatever the llm gives you blind. because the llm are blind too. you cant use them like they're the project lead. they dont know. they can only pretend to know. and pretend they do. you gotta keep jabbin at those infinite monkeys to spit out the right thing. gotta jab carefully. no wrong jabs, or it could land you in a lot of trouble, long time. or even end you short time. that terminator is out there... we believe when you get a machine to do the work of man, you take something away from man. sci-fi quotes^ be tellin it real. and we've got to be honest about it. no hiding in shame. there's no shame in the truth. no problem we cannot mend, but the hidden, the unfound, the denied. so, yes. i've been using llm to expedite this text editor project. had it been up to me alone, i suspect, even if i managed to find the energy to persist, i'd not even be up to feature parity with fin1.001.fish yet. or maybe, generously, fin1.005.fish. and likely more and worse bugs than the llm did. or maybe not. maybe it's easy to tell oneself that, from here, in the blind bias, seeing only the slow start. ... but persist a little, and who knows what pace of learning, what acceleration, of what learned and capable of, and pace of learning to learn too. navigating the fish documentation a lot more, for sure i've atrophied upon, compared to the imagined me who's trying do this without llm. ... but then, the llm ... you know what it's like... the art of prompting. and the gamble of it. roll the dice just right, and you can get all dice on top score. try again. try again. try again. on and on it goes for over 30 rolls back and forth. ... are we sure it's not just better to learn how to do it oneself? ... perhaps both. the better learned how, the better can prompt to expedite some parts. but then one still has to proof read it. and... ... we all know, it's faster to write unreadable code than it is to read it. and, a lot of my hammering at the llm, is to insist on shorter more readable code. ... i've been forgetting to be so verbosely thorough in the opencode interface. feels a bit claustrophobia triggering in opencode, somehow. ... also, just noticed... could integrate it into fin. :3 yeah yeah, sometime way later than ... e.g. syntax highlighting! yes, fin does not even have syntax highlighting yet! okay, to be fair, fin only just got undo/redo. fin's featureset is really still very core minimal. but, speaking of core... that really is the essence of fin, as envisioned. an extensible core. and extensibility's where i've most been leaning to w ... oh, ,there was another bunch of pokes of flaily progress for something i was going to mention, besides syntax highlighting, undo/redo, extensitility... forgot ... most been leaning to extensibility. the prioritisation gubbins keeps suggesting the debuggery stuff. and true. would help. but still thinking, so much is still available to be thrown up into air, into the void, ... the core may still be completely re-written yet. and so, there's much uncertainty around how and where the extensibility fits in. like, am i already going too far... and need to ... idk... modularise extension features, like undo ... heck, even like page up and down? ... backspace and left/right cursor movement wraps lines like how finacky do i go with the extended core model? what even then is the core? the render model? it's clearly janky. heck, even the renderer... does that become a mere extension too? and so... when it gets stripped right back, what's left? just fish? lol. could be very very flexible. fish is already such a more human and enjoyable way to shellscript. yes, that's right. i did say shellscript, not program. because it is shellscript. just not in the purely posix+bashisms way you know it. fish is fish. fish is the friendly interactive shell. interactive features out the wazoo. way more than i know. certainly also way more than i've yet implemented. the interactive features of fish may yet supplant much of the core code so far... so far, here at the progression from a working (and git committed) version of fin, fin1.018.fishsticks.fish, to a fin1.019.fish (prly fishsticks), ... many wrong avenues taken. been trying to do too much still too. not supposed to be doing anything. supposed to be in recovery from burnout. otherwise... y'know the quality of work accomplished in burn-out? so yeah. no stressin about not progressing it, not earning those activity pips on codeberg. rest, rest is not only its own reward, but it'll help recouperate, ready to code up more of fin another day. oh yes, just remembered, the other portion was going to work on, ~ not the remembered the other portion have worked on, ~ the idea of just proceeding through the issues in order. https://codeberg.org/Digit/fin/issues?q=&type=all&sort=oldest&state=open from oldest to newest. which is nice because first, is screenshot, ... a likely rolling open issue, is due an update though (from 0.005 to 0.018, quite different). but then, next real issue is both something was going to work on & something have worked on even did have it working for a while... line numbers up the side. that was way back... had other issues to deal with then, so gutted it out. alas... no recollection exactly where, which version, had that... could likely grep around and find that easy enough. the idea, iirc, was to have line numbers on the side, toggle on Ctrl L. or N. or both. so, after the long wrestling to get undo/redo done, and the still flailing other efforts, line number toggle'd be an easier in reach low hanging fruit ftw. could do with a few easy wins now. hella stuff going on in the world. and that undo/redo shit really kicked my ass. poked me back into burnout. ... "recovery from burnout takes twice as long as you think" says an important post-it note above my monitor. ... so... years yet. years, of staying away from doing 19 hour days again. yeah yeah, my adhdness and autisticness, when clamped on to an inspiration, sure sure, it's great, and can accomplish greatness (~ like writing a text editor in less than a week from scratch, with no prior skill in a way no one else has done before) (or) (~ like designing a house... dont get me started) (or... several other examples). but cant get to any of them, if do not rest and recouperate, from being poked far into burnout. not just this recentmost dip back into it. still recovering from the years long rocket dive into burnout. ... but... it has been nice to have had this poke at something. to see i'm not thoroughly enfeebled for life. still got the something special in me. no that sounds off. was gonna say still got the magic in me. but that sounded off too. still got life in me. :) still got the imaginative innovative spark. ... ... still gonna build the house(s). ... PS dont even have "save as" yet! but yeah... rest. best leave looking at the issues list, until after wellllll rested. https://codeberg.org/Digit/fin/issues PPS oh, just remembered... was gonna mention... loc on latest attempt to implement extensibility refactoring, over 900. so... no. too much for too little. need to whip it into shape. see if it can be done in less than 1 line. (as in refactor for fewer).