Ask HN: How do you use LLMs for private discussions?

9 pointsby emerongi14 uur geleden13 comments

I sometimes have things to discuss with LLMs, which are more private than usual. I was thinking how to do it in a way that doesn't reveal my identity.

  1. Use Tor to access the provider?
  2. Create a random account?
  3. Use some form of untraceable payment (which one?)
  4. Scrub all information provided to the LLM from personally identifiable information?

It seems like a lot of effort. So is running a local LLM, for which I don't even have the hardware. How do you do it?

16 comments

simquat9 minuten geleden

Not sure what kind of model you're interested in, but you may give Duck.ai[0] a try. It doesn't require any signup and let you use gpt-5.4 nano/mini, Haiku 4.5, Mistral Small 4 and Gemma 4 31B.

[0] https://duck.ai/

MyMemoryfails8 uur geleden

Running Local LLM isn't really a lot effort anymore, since its one command away to get the familar interface of chat AI. This setup requires 24GB ish VRAM and 32GB ram, so if you have capable gaming PC which can do AAA games, you can run it.

- llama-server -hf ggml-org/gemma-4-26b-a4b-it-GGUF:Q4_K_M

After that simply open browser and enter: http://localhost:8080

What this does: This will download Gemma4 AI with 26B param & start a http server for chat

Its shockingly capable for its size. Does it beat the top end models? No, but as long your don't fall into the hallucations. Its just fine.

Edit: the software is llama.cpp you can download it from "releases" which u can find at github right side. No need to know how to build it

Edit2: Pro tip is, only use chat per context you want to use. Lots users want "dynamically" change the context, but that doesnt really work from my experience.

740212 uur geleden

With llama.cpp, Intel i9-13900KS CPU, 96 GB RAM, RTX 4070 running locally.

The models I'm using right now with that are:

  gpt-oss-120b-F16.gguf
  Qwen_Qwen3.5-27B-Q4_K_M.gguf
  Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf
  gemma-4-31B-it-UD-Q6_K_XL.gguf

PaiDxng10 uur geleden

The hardest part isn’t payment; the prompt itself may identify you. For truly sensitive topics, I’d abstract the details first, then use a local model—or avoid an LLM entirely.

hbcondo71412 uur geleden

Maybe try Ollama Cloud, Prompt or response data is never logged or trained on:

https://ollama.com/pricing

noman-land9 uur geleden

Best is to use local LLMs obviously. I haven't used it but nano-gpt accepts crypto payments.

https://cake.nano-gpt.com

blinded14 uur geleden

I spin up a gpu instance in a cloud, run my model via vllm, connect to it via an ssh tunnel. done.

stuxnet7913 uur geleden

Can you elaborate on the first step? Which cloud and which service? What's the cost outlay if you are just having a convo and not doing anything 'agentic'?

blinded12 uur geleden

Hey sure.

It depends, but usually spin up an h100 on lambda.ai or coreweave. They have capacity and their UIs/APIs are nice. I spin it up for an hour or two, believe it was 6~ dollars an hour.

Once the gpu instance is up, you need to run vllm and a model, ie https://docs.lambda.ai/education/large-language-models/deplo....

Then you can connect your pi.dev, openwebui, etc etc to vllm and interact with it like normal.

RobotCaleb10 uur geleden

I don't

ahdgs12 uur geleden

i've been using OpenAI's os PII-masking model, works decently and lightweight enough to run virtually anywhere https://huggingface.co/openai/privacy-filter

downbad_14 uur geleden

Can't you do it logged out?

Havoc12 uur geleden

With a local one

HNYN3 uur geleden

makeyouragent8 uur geleden

tomsop10 uur geleden