Back to the channels

#java

15:11 Para Why is Texas limiting nevet?!
15:12 dreamreal ha. ha. ha.
15:13 Para Anything cool going on?
15:15 dreamreal I'm adding an annotation for rss content to nevet so I can train it more gooder
15:15 dreamreal but I found some bugs along the way, as one does
15:16 Para I got a pretty cool maintenance PR in today. -17k rows + 15k rows, most of the diff is package-lock.json going nuts because I moved ancient frontend from whatever webpack to vite+typescript.
15:16 Para Actual diff is like -6k rows of cruft.
15:16 dreamreal heh
15:17 Para And major stuff at that; lots of dao classes to records, more up-to-date and idiomic code all over, testcontainers so that the thing can now actually run the test suite in CI, DevX stuff... all in all pretty neat.
15:18 Para Have to knock on wood, I don't get to spend a week on just making things prettier that often :)
15:22 dreamreal As long as it works!
15:23 dreamreal I've actually been really happy with the UI updates on nevet - I have been ignoring the nextjs UI except for bugs (and since neither I nor anyone else I know uses it: no reported bugs! No reports, no bugs, right?) and the new UIs are both much nicer in completely different ways
15:48 dreamreal https://news.ycombinator.com/item?id=50022246 please upvote if you think it's worth it
15:48 javabot dreamreal's title: "Listening Too Fast | Hacker News"
15:48 nevet Listening Too Fast | Hacker News
15:59 Para From what I understand, there's a specific language processing skill which everyone has and adeptness varies a lot from one person to another; reading and ingesting text without thinking what it _sounds_ like. Speed readers know how to skip the "sound" part entirely, and is the basis of speed reading.
16:00 Para Basically you just need to look at a word to understand it, but something in us wants to push it to the speech part of our brain as well, and that's just extra processing for no reason as you're not going to speak aloud what you're reading.
16:03 dreamreal ... so gimme the upvotes! That's not my karma that you're helping but it IS my site
16:03 dreamreal and did you read the article
16:14 Para I need to re-read it, I'm a terrible reader.
16:15 Para soon :)
16:15 dreamreal I actually pointed that out IN THE SECOND PARAGRAPH!!!
16:15 dreamreal you proved the thesis!
16:16 * dreamreal fumes
16:16 Para Monkey together strong?
16:17 Para Okay, to be serious. I tend to say "only jokes have punchlines", it's basically an inverse of "Put the conclusion first. Don't bury the lede." with the same basic idea.
16:18 dreamreal sure, but ... again, you're addressing the *sender* and not the receiver
16:21 sbalmos dreamreal: oh that's easy. unsafe { receiver = &*sender }.
16:24 Para So I think my particular droning problem is a bit of a difficult one here - for me - to fit into what I think you're trying to say. I blather a lot. My LLM prompts reflect this, it's not that I try to cram a lot of rules and context and whatnot into a prompt, but I end up explaining a lot of tangential things. In human discussions I've gotten better over the years to provide mental checkpoints and
16:24 Para pauses to keep things at least somewhat coherent, but my brain in actuality is still racing way too fast for decent discussion.
16:25 dreamreal Hrmm
16:25 dreamreal so you're missing it, then
16:25 dreamreal which is fair
16:25 Para Most of my prompting strategies nowadays involve "ask first, don't execute" in some form because of this. I've learned to chain myself.
16:25 dreamreal can I still get the upvote though?
16:25 Para I'd have to register.
16:25 Para And I wonder if that would actually be net negative.
16:26 dreamreal well, that's mine, too: I definitely demand explanation and requirement compliance, and I don't let it run tests in isolation ever
16:26 dreamreal it isn't a net negative, no
16:28 Para I really don't understand the sycopanthy in ChatGPT et al, no normal person wants a discussion of any kind work like that.
16:28 Para It used to be absolute garbage a few years ago, it's been tuned down a lot since but it's still kinda horrible.
16:29 dreamreal I think you have a vastly generous opinion of the normal people out there
16:29 Para Fair, my "normal" benchmark is quite far from an average IT worker.
16:29 Para (that'd be my sister who's worked in retail for +20 years)
16:30 dreamreal In my experience there are a lot of people who will use an LLM as inefficiently as possible with absolutely craptastic guardrails, waiting for it to screw up, and then they're like "SEE?!?!?! IT'S ALL A LIE!" ... while someone else using the exact same service and model runs rings around them because they're actually using it properly
16:31 Para Sure. There's actually currently two major LLM driven efforts going which could be used a context. First, the *craft tools, reimplementing Adobe products. The issue tracker is entertaining in its own right, with lofty claims of replacing all Adobe products and all. "Why?" applies there only superficially.
16:32 dreamreal I tryto think generously of my fellow humans but they keep tryna make it hard
16:33 Para The other one is PS5 emulation, which just this week has exploded wildly due to Sony being Sony. There's now _multiple_ emulators, each focusing on slightly different things, with 50...200x performance boosts. One emulator is focusing on building translation layer for the GPU instructions (they just reached 100% coverage), another one is running the PS5 OS, and then there's the more traditional "run
16:33 Para this game" type of emulators also doing leaps.
16:33 Para At the very base level these are the same, reimplement a specific piece of software, but interest and popularity and general acceptance are almost polar opposites.
16:34 Para While people can agree that both Adobe licensing and Sony going digital sucks, somehow essentially the same solution is seen either as "waste of tokens but kinda neat" and "absolutely awesome".
16:35 Para But, it's quite obvious people generally don't listen, no need to use some recent things as example for that.
16:36 Para I had a customer support call (I was calling an ISP) today where I was interrupted because I was apparently explaining enough to the support person's opinion about the rather complex state of circumstances I was having a problem with.
16:36 Para It got kinda frustrating because what I was trying to stay in the very beginning had to be repeated several times a bit later because of said interruption.
16:39 Para And even that is kinda annoying; I recognize that as The Family Resident Nerd I know way too many things beforehand and can provide L2 debugging info right from the get-go, but there's no "skip first level of support" anywhere.
21:24 NeXeN skip-level support, what a great idea! an extension for escalation!