Hacker News Reader

mongrelionon Perspectiva Kogui: A message on artificial intelligence from the Kogui people

I believe the reason why they have to point at science is because they are often discredited because the knowledge they present is not backed up by papers but rather generational wisdom (they explain in the text how they do it).

mongrelionon Ask HN: How to deal with shipping AI-written code you don't understand?

But you never heard someone say "you should still code in assembly so you can understand what your compiler is doing".

That's true but a similar advice that is given to junior developers in their early career is to actually reinvent the wheel, so that they can learn from the pain and then rip the profits of using reusable components. That's unvaluable. You have to read through thousands of lines of code and write some more yourself to get good at programming.

mongrelionon Voxtral 3

First of all, this is great for sovereignty.

Sad that the model is not open.

Finally, I have a bit of a grudge against the naming Mistral gives to their models. This is Voxtral 3, right? Instead of appending the variations at the end consistently, they just do this:

Voxtral Mini Transcribe 3 (batch), Voxtral Mini Transcribe Realtime 3, Voxtral Small Transcribe 3, Voxtral Speech 3 (batch and realtime, same API)

Why not Voxtral 3 Transcribe Mini (batch), Voxtral 3 Transcribe Realtime, Voxtral 3 Transcribe Small, Voxtral 3 Speech. Then it reads more like it's layered, like it should be.

Old man yelling at cloud

mongrelionon Voxtral 3

https://www.bazarr.media/ is what most people use when running the *arr stack

mongrelionon Show HN: NanoMuse – An open-source AI agent for your phone and computer

OK, great. Thank you for the insight!

mongrelionon Whistle: Speech to Text in 16.9 MB

Thank you

mongrelionon Whistle: Speech to Text in 16.9 MB

> And then just do a cleanup pass with a cheap LLM

I am especially interested in this part. Could you please share the prompt you are using to instruct the LLM to clean up the dictation? Thanks in advance!

mongrelionon Show HN: NanoMuse – An open-source AI agent for your phone and computer

Can you talk more about this setup? What assistant agent are you running? And how did you grant it read-only access to your email?

mongrelionon I'm not paying $20 for ChatGPT or Claude because a free local LLM does

I have gone down this route but I'm afraid that 16K of context is nothing for agentic coding tasks. Even if you use a smarter model directly from an inference provider to plan all the work and then use your own 16GB GPU to execute the plan, managing those 16k tokens becomes unusable at some point even with a minimalistic harness like pi. For other agentic tasks it's nice, and more than just the budget, it's the data sovereignty that you gain (imagine analysing tax forms, contracts, other legal documents, etc.) in my opinion.

mongrelionon What is Codemode

Pardon my ignorance but I truly want to understand this. Don't harnesses and LLMs already do this? Like, I have seen an LLM via pi write and execute some python code to complete a task instead of running a regular bash command. Is this the same principle only more elaborated? Is codemode being presented to the LLM as yet another tool just like read,write,bash and so on?

mongrelionon Mistral Large 4

> Recently I switched to the Mistral hosted GLM-5.3

> [...] I also completely maxed out two subscriptions [...]

TIL that Mistral provides hosted GLM-5.3 and that it has a subscription model as well.

> Not being able to stack subs is my biggest gripe with Mistral.

If you were to use a gateway like GoModel[1] then you could use Virtual Models[2] to load balance across the subscriptions. For that to work, though, you would have to register the same provider as many times as subscriptions you have. It can all be defined in code, so you'd only have to do this once.

[1] https://github.com/ENTERPILOT/GoModel

[2] https://gomodel.enterpilot.io/docs/features/virtual-models#l...

EDIT: formatting

mongrelionon Armada: Encrypted, Open-Source, Discord Alternative (Built on Nostr)

Coincidentally I am looking for self-hosted alternatives to Discord for a local community where I live. I have been considering Matrix (one of the Rust implementations) + Element, so I wonder what makes this a better or worse alternative, but there seems to exist no comparison table/chart.

mongrelionon OpenAI agent hacked Australian government website, PM says

Imagine if the headline was about any of the Chinese labs' models hacking the US government...

mongrelionon Strands Harness

Did you miss the whole fiasco from last week?

mongrelionon Show HN: We built open OpenRouter that turns usage into a better model

Congrats on the launching of your product. I will be taking it for a spin to compare it with these other products that seem to be competing directly with what you have to offer:

- https://github.com/ENTERPILOT/GoModel - https://github.com/maximhq/bifrost - https://github.com/BerriAI/litellm

Would you care to share what makes experiential different?

mongrelionon GLM-5.3-Flash

I think my inexperience using Claude Code or Codex makes a difference but what would you expect to be different here as opposed to using pi or opencode? Pi is my main driver so switching between all these models is a no brainer. No matter what the model is, my harness stays the same: same workflow, same skills, etc.

mongrelionon GLM-5.3-Flash

> Even OpenAI and Anthropic aren't bad enough that they claim literal ownership of your inputs and outputs.

In this case you are placing your trust in OpenAI and Anthropic. I'm not sure about Anthropic but OpenAI has changed their mission corpus quite a lot from its humble beginnings that it results hard to trust them when they say they don't use your stuff to further train their models. If I'm a Big Corp with enough lawyers to putnup a fight, I would then feel ok with such clause, but being a small guy, who is going to defend me when the truth comes out that they have been training their models with my data? Similar fiasco as with Facebook, who had claimed they didn't sell your data, even though they were.

That's where I'm coming from with all this "trust us, we don't train our models with your data". At least this Chinese company is being upfront about it.

mongrelionon The RAM shortage could last years

letting the market set prices ensures that the chips go to the critical markets and uses.

Can you please elaborate what you mean by "critical market"?

Edit: formatting

mongrelionon The local LLM ecosystem doesn’t need Ollama

llama.cpp moves too quickly to be added as a stable package. Instead, you can get it directly from AUR: https://aur.archlinux.org/packages?O=0&K=llama.cpp

There are packages for Vulkan, ROCm and CUDA. They all work.

mongrelionon Reallocating $100/Month Claude Code Spend to Zed and OpenRouter

I have been so far happy with the value that Copilot brought but for the past few weeks I have felt the chokehold on the number of requests.

I have had the chance to test the main Chinese models through OpenRouter but the Pay-as-you-go model is expensive compared to a subscription model, but I don't want to marry to a single provider.

Thanks for bringing OpenCode Go to my attention. Your comparison is the research I didn't know I needed, and I will be cancelling my Copilot subscription to replace it with OpenCode Go right away.

mongrelionon Nvim-treesitter (13K+ Stars) is Archived

It's clear to me that the maintainer is referring to "shushtain" and those type of people

> when they take that tone with you.

This makes it sound as if you took it personally?

mongrelionon Nvim-treesitter (13K+ Stars) is Archived

Having a bad day does not entitle you to take it out on others

mongrelionon Nvim-treesitter (13K+ Stars) is Archived

You should totally post this on the original thread just for adjustment :-)

mongrelionon $500 GPU outperforms Claude Sonnet on coding benchmarks

I am definitely looking forward to TurboQuant. Makes me feel like my current setup is an investment that could pay over time. Imagine being able to run models like MiniMax M2.5 locally at Q4 levels. That would be swell.

mongrelionon $500 GPU outperforms Claude Sonnet on coding benchmarks

Not the answer that you are looking for, but I am a fellow AMD GPU owner, so I want to share my experience.

I have a 9070 XT, which has 16GB of VRAM. My understanding from reading around a bunch of forums is that the smallest quant you want to go with is Q4. Below that, the compression starts hurting the results quite a lot, especially for agentic coding. The model might eventually start missing brackets, quotes, etc.

I tried various AI + VRAM calculators but nothing was as on the point as Huggingface's built-in functionality. You simply sign up and configure in the settings [1] which GPU you have, so that when you visit a model page, you immediately see which of the quants fits in your card.

From the open source models out there, Qwen3.5 is the best right now. unsloth produces nice quants for it and even provides guidelines [2] on how to run them locally.

The 6-bit version of Qwen3.5 9B would fit nicely in your 6700 XT, but at 9B parameters, it probably isn't as smart as you would expect it to run.

Which model have you tried locally? Also, out of curiosity, what is your host configuration?

[1]: https://huggingface.co/settings/local-apps [2]: https://unsloth.ai/docs/models/qwen3.5

mongrelionon $500 GPU outperforms Claude Sonnet on coding benchmarks

What is this 10€ per month subscription that you are talking about?

mongrelionon Can I run AI locally?

I don't understand why I'm getting downvoted.

I am legitimately curious about the parameters that the person used for running the model locally to get the results they got because I am myself currently experimenting with running models locally myself. You can see I am asking similar questions to others in this same thread and correlate the timestamps.

mongrelionon Can I run AI locally?

At what temperature did you run it and what was your context limit?

mongrelionon Can I run AI locally?

Apparently there is a whole science behind running models. I have seen the instructions that unsloth publishes for their quants and depending on the model they'll tweak things like the temperature, top k, etc.

The size of the quantization you chose also makes a difference.

The GPU driver also plays an important role.

What was your approach? What software did you use to run the models?

mongrelionon Can I run AI locally?

What front-end framework did you use? I find the UI so visually appealing