HNHacker News
TopNewBestAskShowJobs

fzxu22

27 karma · joined January 22, 2022

submissionscomments
fzxu22··on Ask HN: What's your current AI agentic setup?
1. For heavier works im using deepseek v4.1 flash and pro, I also have z.ai coding subscription but probably trying to move away from all of these because:

2. I just got a M5 Max 128G mac studio and self-hosting models now. I did some benchmarks on some potential setups, but currently I’m settling with the Qwen series models (3.8 27B/Flash Next for better intelligence, 3.5 35B A3B for spead), serving with MTPLX. Compared to API providers they are slower and has less context length, but enough for most of my use cases. I am exploring routing solutions between self-hosted and API providers.

3. Mostly Pi (oh-my-pi for main coding env, bare bone pi for drop in usages). Its small footprint is really important for self hosted models

4. I don’t think so, I do have a nanoclaw setup a while ago and am trying to port it to use pi as driver and connect to self hosted models.

5. No, they’re never to be trusted.

fzxu22··on Talorys – A self-hosted personal AI agent on Cloudflare's free tier
Please don't use the word "self-hosted" if it's not hosted on your hardware. This thing depends entirely on cloudflares services and "generosity".
fzxu22··on Auto mode is now the default in Claude Code
Want to use this thread to ask the wise folks on HN: I’ve been running auto mode but always run within a docker container with only the necessary workspace dir mounted in. Is that considered safe or actually not?
fzxu22··on Don't use your phone while you poop
I poop for like 30 seconds so not very useful
fzxu22··on Anthropic's model naming, extrapolated
Can't wait to use Claude Cinematic Universe!
fzxu22··on Codex just found a "workaround" of not having sudo on my PC
That’s why I only run coding agents in containers with limited access to host resources.
fzxu22··on OpenAI Privacy Filter
Thanks for the comment! And yes, redaction based measures will always face the trade-off between privacy vs intellegence you get out of LLM. e.g. provider will inevitably know you're in some sorts of legal/tax issue even without any PII. And for some case the intellegence you want will depend on LLM knowing some detail (e.g. your AGI when doing tax preparation).

On the other hand Moxie's Confer is really interesting! On first glance I thought it's using homomorphic encryption but it turns out to be based on hardware isolation. TIL +1

fzxu22··on OpenAI Privacy Filter
The proxy keeps 2-way mapping of identified PII and the redaction e.g. Jane Doe <-> <PERSON_1> so the process is reversable i.e. redactions from LLM response will be replaced back to the original, and it should feel transparent on user end. I'll add more detailed example in README to make it clear.
fzxu22··on OpenAI Privacy Filter
Working on this: https://github.com/KevinXuxuxu/anon_proxy, a sort of anonymization proxy to use with LLM providers. It does model (OpenAI privacy filter) + regex PII detection, and replaces them back-and-forth for API requests and responses. With locally hosted detection model, no PII leaves your local environment. I find it very useful especially when you're working on sensitive documents (legal, tax, immigration etc.), hope you find it helpful as well :)
fzxu22··on Ask HN: Share your personal website
https://site.fzxu.me

The index is a bit outdated, it’s mostly about me working on my homelab.

fzxu22··on RSS is awesome
+1 happy freshrss user here
fzxu22··on Launch HN: Airhart Aeronautics (YC S22) – A modern personal airplane
Getting cheaper and better planes is good. But this is not a correct way of solving traffic problems in US. We need high-speed railroad network, not one airplane per household.
fzxu22··on Show HN: Simple Wordle solver in command line in Python
LOL that's actually very useful as I'm trying to find out how the thing works
fzxu22··on Show HN: Simple Wordle solver in command line in Python
Probably true according to this post: https://slc.is/#Best%20Wordle%20Strategy%20%E2%80%94%20Explo...
fzxu22··on Show HN: Simple Wordle solver in command line in Python
You ... are not wrong ;)
fzxu22··on Show HN: Simple Wordle solver in command line in Python
Thanks for the reply! I really appreciate your feedback on this. I wasn't giving much thought about coding style and naming etc. when I worked on this, just wanted to get it running asap. I'll definitely take a better look at your suggestion (I'm not very familiar with perl) and see if I find any time to improve this thing. cheers.
fzxu22··on Show HN: Simple Wordle solver in command line in Python
see README