#test | Logs for 2025-05-15
« return
[03:29:12] <chromas> =y test
[03:29:15] <systemd> Search failed: Could not connect to server on handle 5C74604FA100
[04:00:00] -!- systemd has quit [Remote host closed the connection]
[04:00:09] -!- systemd [systemd!~systemd@50.54.gin.jp] has joined #test
[04:00:09] -!- systemd has quit [Changing host]
[04:00:09] -!- systemd [systemd!~systemd@pid1] has joined #test
[04:00:09] -!- mode/#test [+v systemd] by Artemis
[04:00:50] <chromas> =y kanye nhh
[04:00:53] <systemd> Search failed: Could not connect to server on handle 574CF044BE70
[04:00:58] <chromas> waaaaaat
[04:01:10] <janrinok> mornin' chromas
[04:01:16] <chromas> g'day
[04:01:27] <chromas> Finally got yacy search back up
[04:01:31] <chromas> the bot disagrees
[04:01:36] <janrinok> lol
[04:02:27] -!- systemd has quit [Remote host closed the connection]
[04:04:32] <chromas> it's somehow got an ancient build
[04:04:42] <chromas> oh yeah I'm retahded
[04:05:39] -!- systemd [systemd!~systemd@50.54.gin.jp] has joined #test
[04:05:39] -!- systemd has quit [Changing host]
[04:05:39] -!- systemd [systemd!~systemd@pid1] has joined #test
[04:05:39] -!- mode/#test [+v systemd] by Artemis
[04:05:58] <chromas> =y kanye nhh
[04:06:08] <systemd> http://norvig.com - count_1w.txt
[04:06:46] <janrinok> I don't think it is meant to show us that, is it?
[04:06:53] <chromas> no idea why that's the top result, but the guy's got some stuff on his site
[04:07:11] <chromas> Peter Norvig of google research
[04:07:39] <chromas> more importantly, I can turn on yacy backend in searxng
[04:08:53] <chromas> and while I'm at it, submitting articles should fail less now
[04:08:57] <janrinok> I don't recognise the 'yacy' name
[04:09:07] <chromas> it's a p2p web search system
[04:09:16] <janrinok> ah OK
[04:09:52] <chromas> and searxng is a meta search engine, which queries others like google and whoever, so you get a bit less tracking (since only the server connects to google and whatnot)
[04:10:45] <chromas> They used to support yandex, which is great for bypassing american filtration, but it's an anti-bot-ite now
[04:12:25] <chromas> So now the bot's got Google, DDG, Yacy and SearXNG searches
[04:12:29] <chromas> I think it's got DDG
[04:12:32] <chromas> =d bewbs
[04:12:37] <chromas> =ddg bewbs
[04:12:37] <systemd> https://www.noslang.com BEWBS Meaning & Definition
[04:18:30] <janrinok> it's got something...
[04:30:14] <janrinok> so, we have to specify the search engine that we want to use each time, or is there a default?
[04:31:39] <janrinok> =y ECCM
[04:31:41] <systemd> https://aurora2.pentarch.org - Potential ECM/ECCM Changes
[04:32:20] <chromas> different lettres
[04:32:28] <chromas> =g =y =ddg =s
[04:32:28] <systemd> https://pubmed.ncbi.nlm.nih.gov - Membrane topology analysis of the Escherichia coli aromatic amino ...
[04:33:14] <chromas> o' course, using searx through the bot doesn't give you any privacy enhancements, since you're already not connecting out :)
[04:33:16] <janrinok> oooh - french spelling of lettres too
[04:33:34] <chromas> who proxies the proxies?
[04:35:00] <chromas> yacy search results are pretty garbage. Guess I need to build up my own index
[04:36:21] <janrinok> do you cache anything?
[04:36:46] <chromas> I think it does
[04:37:16] <chromas> I can have it crawl pages and index them, but it also queries other people's instances and that seems to be where it has trouble getting decent results
[04:37:33] <chromas> You can also run an isolated instance, like if you just want search for your own intranet or whatever
[04:38:20] <chromas> I'm just trying to get better results with less reliance on goog and bong
[04:38:29] <janrinok> I searched for ECCM but the best that it could come up with was a game that was pretending to have ECCM when spaceships attack 'Aurora'.
[04:38:49] <janrinok> Yeah, I get that, and it is a worthy attempt too
[04:39:03] <janrinok> goog is next to useless for me now
[04:39:26] <janrinok> ... unless I search down several pages to find genuine results
[04:39:45] <janrinok> I can't be arsed usually, so I go elsewhere
[04:40:13] <janrinok> =g ECCM
[04:40:13] <systemd> https://www.eccm.org - ECCM: Expert County Care and Crisis Services in Erie, PA
[04:40:26] <janrinok> that is even worse that the last search!
[04:41:45] <chromas> hm I thought it was the x11 protocol extension
[04:41:47] <chromas> =s eccm
[04:41:48] <systemd> Electronic counter-countermeasure - Wikipedia - https://en.wikipedia.org
[04:42:33] <chromas> =ddg eccm
[04:42:34] <systemd> https://en.wikipedia.org Electronic counter-countermeasure - Wikipedia
[04:42:42] <chromas> =y eccm
[04:42:43] <systemd> http://annotate.de - ECCM
[04:44:19] <janrinok> -g Airborne ECCM - Poseidon MRA1
[04:44:58] <janrinok> oops, it doesn't seem to like that
[04:45:09] <janrinok> -y Airborne ECCM - Poseidon MRA1
[04:45:13] <chromas> it's probably the -
[04:45:20] <janrinok> -y Airborne ECCM Poseidon MRA1
[04:45:26] <chromas> the first one
[04:45:32] <chromas> you're looking for =
[04:45:33] <janrinok> -g Airborne ECCM Poseidon MRA1
[04:45:51] <janrinok> nah, there is not much about it.
[04:46:01] <chromas> =y Airborne ECCM - Poseidon MRA1
[04:46:03] <janrinok> -d Airborne ECCM Poseidon MRA1
[04:46:04] <systemd> [0 results]
[04:46:11] <chromas> =s Airborne ECCM - Poseidon MRA1
[04:46:12] <systemd> Boeing P-8 Poseidon - Wikipedia - https://en.wikipedia.org
[04:46:40] <janrinok> at least I got a [0 results] that time
[04:46:52] <chromas> I did. You got nothin' :D
[04:47:05] <janrinok> lol
[04:51:42] <chromas> startpage is still around, if you want google without google
[05:00:05] <janrinok> At least startpage found some relevant articles - although a bit dated.
[05:07:21] <chromas> We need to get someone to start a Rustify Yacy project
[05:09:50] <janrinok> what language is it written in?
[05:11:34] <chromas> java
[05:11:38] <janrinok> lol
[05:11:47] <janrinok> write once - run anywhere?
[05:12:05] <chromas> as long as you have the right jve version
[05:12:16] <chromas> looks like current ones work now. I think it was stuck on 8 for a long time
[05:12:36] <chromas> Nice thing about java though is the support for everything, maybe even more than python
[05:13:12] <chromas> support for more database system than there are stars in the sky
[05:13:51] <janrinok> I might look at it again - not that I write in java! But there were some useful bits of software around but they always required that I install some out-of-date software to get them to run. That might be an OS problem though.
[05:14:10] <kolie2> o/
[05:14:33] <janrinok> I haven't looked at it in several years, but my GO and Python IDEs are written in it apparently!
[05:14:33] <chromas> heh, maybe we need an instance for searching SN :D
[05:14:43] <chromas> yep
[05:14:44] <janrinok> kolie \o
[05:14:58] <janrinok> kolie2, \o
[05:14:59] <chromas> jetbrains is all about java
[05:15:15] <kolie2> They make good dotnet apps too.
[05:15:19] <janrinok> they work quite well, I am happy with them
[05:15:24] <chromas> in fact, google switched the android studio from fartbox or whatever the opensource thing is to jetbrains
[05:15:33] <kolie2> VSCode is where its at for sure. Visual Studio 202X is solid af aswell.
[05:15:34] <chromas> (that was years ago now though)
[05:15:50] <kolie2> python in vscode is chefs kiss.
[05:17:14] <janrinok> I'll stick to my jetbrains IDE's - they are all identical in operation and function, and I have a professional licence now. Well worth it for me.
[05:17:31] <chromas> does it give you the database manager?
[05:17:36] <janrinok> yep
[05:17:39] <chromas> they sell it separately as DataGrip
[05:17:46] <janrinok> and containers, and .....
[05:17:52] <chromas> Looks pretty sweet, but I went with DBeaver which is good enough for now
[05:17:59] <chromas> =yt weird al good enough for now
[05:18:00] <kolie2> jetbrains is very nice.
[05:18:00] <systemd> https://youtube.com - Good Enough for Now (03:04; 119,386 views; 👍1,842)
[05:18:16] <kolie2> pycharm?
[05:18:41] <janrinok> yep and goland
[05:20:16] <kolie2> my problem with jetbrains is the ai integration is well, super basic, and all the really advanced stuff is only in vscode or AI specific ides ( aider etc )
[05:20:30] <kolie2> Other then that, straight coding, they are top notch.
[05:20:47] <janrinok> I don't use any AI so it is disabled as part of my config.
[05:22:31] <kolie2> Yea what usually ends up happening is, you have your favorite IDE, and then the same workspace open in the other ai efficient tool.
[05:23:07] <chromas> does that give you more features than using an lsp?
[05:23:58] <kolie2> I mean the LSP let's the ide do the refactor and code completion and stuff.
[05:24:18] <kolie2> I think AI in general has taken little advantage of LSP type info rn, and that's a major way to boost it.
[05:24:32] <janrinok> You get that whether AI is enabled or not.
[05:24:35] <kolie2> like using a pdb to understand types and classes would be ahead of it.
[05:25:06] <kolie2> AI's have access to MCPs now, you could create a MCP=>LSP interface and then provide prompts which guide LSP usage for useful insights.
[05:25:10] <chromas> ah, I just know people are integrating the ais with the lsp interface, like into neovim and stuff
[05:25:52] <kolie2> VSCode Copilot can access the lsp.
[05:25:58] <kolie2> I think cline does as well.
[05:26:09] <kolie2> I'm using vscode with a cline spinoff.
[05:26:25] <kolie2> https://github.com
[05:26:26] <systemd> ^ 03GitHub - SilasMarvin/lsp-ai: LSP-AI is an open-source language server that serves as a backend for AI-powered functionality, designed to assist and empower software engineers, not replace them.
[05:27:04] <kolie2> With LSP you can give it general macro style tasks, code refactoring etc, that are custom actions.
[05:27:41] <janrinok> I like the fact that GO can cross-compile code too. I get things working on my main computers, cross-compile to ARM, and they run immediately on my Pi
[05:28:19] <kolie2> What I've been working on is AIs that break down tasks, and issue responses to themselves, and basic iterate on their own output to perform a complex multi step function. I've been running one right now that's called itself roughly 15000 times.
[05:28:38] <chromas> recursive thinking stage
[05:28:52] <chromas> can LLM cross-compile?
[05:28:58] <janrinok> be careful, you could end up disappearing under your own exhaust!
[05:29:42] <janrinok> chromas, I wouldn't even try - it takes seconds in GO
[05:29:51] <kolie2> The LLM models can be hosted on a variety of platforms, they are basically a weighted number set, you need a program that runs on that architecture, to load the numbers.
[05:30:10] <kolie2> all major archs have LLM loaders.
[05:30:25] <chromas> yeah but you need a gpu with bigass overpriced vram is you want any speed
[05:30:27] <kolie2> The really useful stuff needs 8x 24gb vram cards in parallel.
[05:30:52] <kolie2> I run local models on 16gb ram, and they are ok for some things, when it comes to coding though, stick with cloud models.
[05:30:53] <janrinok> lol - I'll reread my Rasp Pi specs....
[05:31:14] <kolie2> sonnet 3.7 or gemini 2.5 pro is basically god mode for any task at all.
[05:31:36] <kolie2> I run some models on my rpi.
[05:32:00] <janrinok> you must have tons of patience...
[05:32:00] <kolie2> spam, sentiment anaylsis, tool use, function level coding tasks.
[05:32:19] <kolie2> depends on the model, you can get some okay token rates.
[05:32:38] <janrinok> Several of my Pi's are remote so I don't want to have to be changing them frequently.
[05:32:59] <kolie2> yea i have a rack in a DC with 4 rpi 5s rn
[05:33:18] <kolie2> Gemma, Llama3, mistral, phi2 run on pi.
[05:33:22] <kolie2> Q4 7b models run well.
[05:34:29] <chromas> where's our dollar store tensor processors with slots for 480GB of ram?
[05:34:48] <janrinok> still in our dreams apparently
[05:35:30] <janrinok> anyway we have 'Pound Stores' which sound like much more fun, but aren't
[05:36:43] * janrinok is referring to UK of course, not sure what the French equivalent would be...
[05:37:11] <chromas> omelette du fromage
[05:37:20] <janrinok> lol - not quite!
[05:38:26] <kolie2> night dudes.
[05:38:56] <janrinok> g'night kolie2
[05:39:09] <janrinok> see you later on today
[05:39:12] <chromas> g'noight mang
[06:25:58] <chromas> so far the yacy results are few and retarded
[06:26:38] <chromas> it's like seeing a montage of people on the street answering some question, then cutting to the Matt Daemon doll from Team America
[06:30:58] <chromas> not enough memory. java, goddamnit
[06:41:08] <janrinok> lol
[06:41:46] <janrinok> "640Kb should be enough for anybody"
[06:41:48] <chromas> set it to use 8 jiggabytes, it died trying to restart itself
[06:42:37] <janrinok> wait and see what happens if you move it at 80 MPH!
[06:43:09] <chromas> 88 Gotta make sure all the segments of the LED display are lit up for maximum flux capacitance
[06:43:26] <chromas> I think the trick is to locally index anything that might be interesting, so that good stuff shows up in the results