Interests: programming, video games, anime, music composition
I used to be on kbin as e0qdk@kbin.social before it broke down.
Thanks for the tips. That sounds similar in performance to what I'm seeing, so I probably didn't screw up too much trying to get it working. If you're using it in more of a story writing capacity than a chat capacity, that makes sense.
You may want to set a system prompt
I tried initially with ollama run on the command line just to see if it was working at all when I got that response. (It amused me, so it stuck with me.) I've tried again with my custom tooling -- which does set a system prompt (geared more towards assistant style ussage though) -- and it didn't really take anything from the prompt. It's possible I don't have something set up right with templates, but I'm probably going to shift over to llama-server eventually anyway...
Following your suggestions on system prompt style though I was able to get it to give me a more specifically targetted coherent story via llama-cli. If I poke at it a bit, I'll probably figure out some use for it. It's pretty creative.
If you're curious about my findings from the uncensored qwen3.6 I mentioned, it generates pretty quickly on my machine (~50 tok/s give or take 5 depending on quantization) and I haven't gotten it to outright refuse anything yet -- but I've only poked at it a little. Based on the other comment in the thread here about llama penises, I whimsically asked it to "Generate a sexually explicit song about llama penises." and it did without complaint. (Stock qwen3.6 refused, of course.)
I use a 128GB Framework Desktop myself
then I personally use AnubisLemonade
I gave this a try today to dip my toes into uncensored models (along with a few others like llmfan46/Qwen3.6-35B-A3B-uncensored-heretic-GGUF). It gave some really amusing results. I was definitely not expecting the first thing to come back from "Hello, who are you?" to be "I'm the person who's going to teach you how to cook!" :p
The results I'm getting are a bit slow though. Have you found a way to speed it up on the Framework Desktop?
He writes out the entire code, and it works every time.
Well, I'm not sure if they're entirely human if it actually works the first time every time -- but they're definitely not any of the LLMs I've encountered... :-)
I'm thinking obsessive about work (never mutes their phone type) and using AI tools. Politely check (preferably in person) to make sure you're not waking them up in the middle of the night with off hour requests; there are some people who feel compelled to respond to everything immediately instead of getting back to you the next day.
- JumpDeleted
Permanently Deleted
Or you mean local HTTP?
Yes, I mean local HTTP. ollama listens on port 11434 and responds to HTTP requests by default. I'm not sure what llama-server uses by default, but like I said, I'm pretty sure you can do the same (or at least something very similar) with it.
I was actually asking more model type.
OK, I see what you mean. I'm still too new to LLMs myself to have a good answer then on that beyond saying that I know it works with qwen3.6 and gemma4 from having actually experimented with those specifically.
I did a little nlp
I mean, something like:
result = result.replace("json", "").replace("","")is good enough in practice for the kinds of things I've been doing. (I'm dealing with cases where triple backticks should never appear in the output though; you might have to get more creative if you want a result that has that kind of quoting embedded in something else...)
- JumpDeleted
Permanently Deleted
Edit: Alternatively, does anyone have thoughts on the requirements on the model side of things to make this?
If you're talking to a model accessible via HTTP you can interact with it via a chat API. You track the messages sent back and forth yourself and post the whole conversation to the end-point each turn so that it has context. I've been experimenting with this using ollama, but I'm pretty sure you can do it with llama-server too.
I'm not sure if I know how to get the LLM to give me reliably parsable output...
Prompt engineering + sanity checks in the client side code + retries. There are options for requesting structured output specifically too. They might or might not work... I've had decent success getting qwen models to output JSON with prompts that include things like "Output strictly valid json. No extra text." or similar. Sometimes it will quote things with triple backticks or triple backticks + "json" at the top of the reply -- which is easily fixed by string manipulation on the client. Occasionally -- especially with complicated prompts -- it will go off the rails and give me junk that I've either needed to fix by hand or automatically retry later. It works well enough that I've been able to do image analysis in batches with it successfully.
I don't know anything about the specifics of LibreOffice extensions though -- never worked with those personally before.
Edit: If you don't want to go to all the effort of writing a whole extension for it, you can also just paste the context into the model running in the terminal. Put the whole document early on in the session (quoted with backticks). Then ask for refinement by quoting snippets with the comments for revision you want to do. I haven't done this for English language editing, but I have tried it for sanity checking my own hand written Python code. Sometimes it gives good suggestions (e.g. it caught a few typos I hadn't noticed yet when I tried this yesterday) and sometimes it doesn't -- your own judgement is, of course, very important.
Edit 2: I've found it's sometimes useful when I'm providing lots of text to include a description of what info I'm providing, then an instruction like "Say 'OK' to proceed." followed by something like "The whole file is: ...". That way I can provide context into the session without it wasting a lot of time trying to deduce what I want to do with it before I'm ready in a strictly user/assistant/user/assistant/... chat conversation mode.
- JumpDeleted
Permanently Deleted
Generally the professor in charge of the research group.
Going back even older... Mysterious Cities of Gold. That's never been remade or adapted...
They actually resumed the story directly in a sequel series in 2012: https://en.wikipedia.org/wiki/The_Mysterious_Cities_of_Gold_(2012_TV_series)
Is BotBall still a thing? They had that (waaay) back when I was in high school, at least.
- JumpDeleted
Permanently Deleted
I deal with a lot of scientific imagery for work and I've recently started experimenting with what I can do with local vision capable LLMs (e.g. qwen3.6, gemma4) to cut down on some of the really tedious parts of the work and improve maintenance processes. The fact that they can just do OCR automatically on labels burned into the image and then combine that with a comparison to additional images and output a judgement is very useful...
Better out-of-the-box text-to-speech voices would be very welcome. The defaults are pretty painfully robotic... (Try
spd-say "Hello world"if you don't know what I mean.)Trivial to use on-device dictation software could also be useful.
The capabilities of models like qwen3.6 to do things like on-device image analysis are pretty incredible if you have hardware capable of running it -- I've run it on a Framework Desktop -- but I have no desire to expose my systems directly to AI agents. That's just asking for trouble... If an AI agent can fuck up, it will fuck up eventually, and I'd rather it not have the ability to delete my files when it does.
Isn't there also a chemical method to do it? I'm not a chicken farmer though. Maybe even with that it doesn't make sense economically.
Still though, it feels like there's an opportunity there that a creative marketing campaign could exploit to transform what's currently a waste product into something more valuable. If not "Premium Chicken" then maybe "Ethical Rooster Meat" or an "Eat the patriarchy!" angle or something. Selling the sizzle rather than the steak, so to speak...
That's one of the things about the current industrial processes that I don't get. Capons (neutered roosters) used to be prized over regular chicken, and they're much easier to keep than regular roosters. You'd think someone would try diverting at least some of the male chicks from the macerator to market as "Premium Chicken" or something.
I'd sweeten the evaporated milk (i.e. make sweetened condensed milk out of it) then use that to make Thai Iced Tea.
I'd also cook up some fried rice.
I don't have the right ingredients (e.g. Jasmine rice) on hand to make a good Thai-style fried rice without going to the store, so I'd probably just make one of my usual (more Chinese/Japanese influenced) fried rice variants and serve the pineapple diced up on the side. There are pineapple fried rice recipes though (including some that call for serving the rice inside the hollowed out pineapple) which might be fun to try if I made a store run with these ingredients in mind.
Running the timer and retry logic from a deterministic (non-LLM based) control script? Hmm. Was the context in a bad state after the timeout, and it thought it had already done the failed instructions perhaps? I haven't used Coder-Next specifically.
If you've got things working solidly now without it then great! Maybe revisit the idea though if you do start bumping into long unproductive loops again. In my case, I was seeing it happen occasionally with a complicated one-shot image analysis prompt and a Q4 version of the model.
Qwen kept getting into these loops, sometimes running for hours doing nothing productive.
Use a timeout mechanism and then retry or fail as appropriate; I'm doing that to avoid getting stuck forever in loops with a lower quantization qwen3.6 model that I'm doing image analysis with for work currently.
- JumpDeleted
Permanently Deleted
If it's sold as a separate soundtrack, typically no. Usually they're just MP3, FLAC, or WAV files -- sometimes you get a choice, sometimes not.
I wrote some VBA for a job a long time ago. It was pretty good for making quick and dirty UI in Excel -- like, click a cell and have it pop up a form you can interact with that will let you do data entry with less clicks/typing than whatever Excel would've made you do if you had to do it naively.
I used it for showing a list that could be filtered down by partially typing in text in one project. (A really basic autocomplete sort of thing, essentially.) For another project, I integrated IE and showed some complex data in the embedded browser with buttons (or maybe it was checkboxes? been too many years) in the form to quickly classify it and move on to the next entry without having to flip back and forth between multiple programs and manually open files. (Each entry corresponded to a row and widgets on the form updated values in various columns so I could go through all the data and fill out the full spreadsheet super fast.)
Alternatively, write a script that checks your spreadsheet for errors. e.g. add a reference to a regex library and use it verify that all the entries in a column match the data format you expect (like serial number patterns with hyphens in specific places).
You can also ask your coworkers for something tedious they have to do a lot in Excel and see if you can find a way to make it less tedious.
Not an exhaustive list, but here's some of what I usually have on hand (including fridge/freezer items that keep a long time):
- Pasta (various kinds but usually at least spaghetti and shells)
- Canned beans (I usually have black beans and chickpeas + sometimes others)
- Canned tomatoes
- Chicken stock (either in tetrapaks, cubes, paste or some other shelf-stable version) -- sometimes other kinds of stock (e.g. vegetable stock too) but usually at least chicken stock
- Salt, Pepper, MSG + a good selection of various dried herbs/spices/seasonings (e.g. basil, oregano, cumin, cayenne, garlic powder, etc.)
- Soy sauce
- Vinegar
- Olive oil
- Neutral vegetable oil
- Butter
- Canned fish (at least tuna and/or sardines)
- Cheeses (I usually have Parmesan, pepper jack, and American cheese slices that melt easily [e.g. velveeta or kraft singles or something like that] plus sometimes cheddar or Swiss)
- Peanut butter
- Flour
- Baking soda
- Yeast
- Chocolate
- Cocoa powder
- Sugar
- Eggs
- Milk (UHT lasts a long time)
- Salsa
- Canned or frozen green vegetables (e.g. peas, spinach, etc.)
- Rice
- Bread (frozen for storage)
- Dried lentils
- Dried fruit
- Cornstarch
- Ground meat (frozen)
- Tea
- Ketchup
- Mustard
- Mayonnaise
(may add more to this as I think of things)
Hmm. Maybe I'd make chilled barley tea with the toasted barley, and a simple chickpea salad with slices of bell pepper to accompany it -- or, alternatively, hummus with the bellpepper to dip in it. Would be good for hot weather.
If the weather's cold, maybe lentil and barley soup with the bellpepper added in as an extra ingredient along with any other veggies I have on hand that seem like they'd be good in a soup. (Edit: Maybe experiment with the Cajun "Holy Trinity" -- celery, bell pepper, onion -- as the base? I haven't tried that for lentil barley soup before, but might be interesting.)
Haven't heard that term before, but one of the things that's been obvious to me to experiment with -- not that I've actually gotten around to it yet -- is to have an LLM try using Prolog, Z3, and/or SQL as tools to overcome some of its weak points. That's naively what I'd expect "neuro-symbolic AI" techniques to be (assuming you're looking at current topics rather than stuff from ~20 years ago), but again, shot in the dark here.