

Cuck coding an anti-cheat should be interesting. I wonder how many “you are right to push back, I left a giant vulnerability open that can be exploited by fornight, let me fix that for you” they will come across.


Cuck coding an anti-cheat should be interesting. I wonder how many “you are right to push back, I left a giant vulnerability open that can be exploited by fornight, let me fix that for you” they will come across.


The multi language is great, true, but that starts with good multi language STT, which in my experience, Flemish Dutch has like a 70-80% error rate because they are all trained on Netherlands dutch, and that isn’t even with a west Flemish accent lol.


I am confused. The default whisper + HomeAssistant assistant + piper pipeline works completely fine for day to day tasks, turning lights on and off, etc…
Why would OP try to host a local LLM for the same basic intent recognition that home assistant does pretty well already?


I need to do something similar. I have a Esp32S3 music player/ voice assistant connected to my old Yamaha, 90s AV receiver. It can change the volume knob via IR commands (cool integrated motor that physically turns the knob when the remote or HA changes the volume with IR).
I want to do something like put a ring magnet and an encoder on the tiny shaft or something to get position feedback to couple to HomeAssistant/musicassistant.


I read the whole thing. This seems like standard sponsored response of not biting the hand that feeds them now that they are full time…
The entire point 3 of the user complaint is reaching really far. The one user is being toxic through the whole thread so I also get from their side not wanting to heavily engage because nothing besides divebombing immich’s entire livelihood would satisfy them it seems.
FUTO seems like a toxic company, but the thread isn’t damning. It seems like we have to keep an eye on immich. Otherwise a fork I guess…


FUTO sponsors immich I guess, FUTO is on there. FUTO is problematic but as far as anyone can tell, they haven’t pushed anything on immich.


It was a joke, chill.
But let’s be in good faith about this, vibe coders will overwhelmingly use cloud based corporate frontier models…
And self hosting LLMs in general singularly and only solves one part of the resource objection (prompts running on green energy) and that is all and only if the model is run on green energy. It doesn’t solve most of the moral objections, IP objections, economic objections, resource objections, or brain damage objections.
That being said, one’s own, local, never-to-release app or internal tooling is probably one of the better uses for LLMs like documentation discovery, informal document translation and tiny TTS and STT models (the best usecase in my own opinion)


“You’re right to push back, when you create a new plant layout on Fridays, it does drop the entire database and rm -rf / your system, would you like me to fix that for you?”


Look at the code and commit history and/or talk to the developer.
If the developer only commits in giant blocks of 1-2 days and mostly in a few hours and changes thousands of lines of code in that time, the chances of it not being vibe coded and actually being checked are extremely slim, outside of a complete project refactor. It used to be that LLMs would just do massive single commits, but now they do a thousand tiny commits on the same day using a ton of barely human-readable jargon (for less experienced devs) in the git commits.
If the project was started past 2025 and the initial commit(s) on the first day were massive, that also points (less strongly) towards vibe coding.
If the Dev can’t explain what his code does without having an LLM write the response (this is getting to be harder to spot because LLMs are good at language patterns), that also generally also means that it was vibe coded.
If the dev’s commit history only includes documentation, issues, or very very small things until 2025-2026 and combined with one of the above, it is usually vibe coded.
If the dev has been writing issues or commits jn AI tool repos, that is another indicator.
The vibe coding identifying sites online are complete garbage with mostly false positives. For example, if I run it on my old github repos from before LLM coding existed, it says that it is very likely vibe coded because I comment a ton because my memory isn’t great and I am an electronics guy with very bad coding git management so my git commit history is like one or two commits per day when I am working on something, but I try to bullet list all of my changes in a readable way for people not familiar with the code (like myself in 1 year lol)


I think the comment about AI is not in good faith.
It isn’t users saying every “AI assisted” project is vibe coded slop.
It is just as often or more that developers and “developers” lying about AI assistance when they are just vibe coding or “AI assisting” and barely checking over the code. There are tons if those on Fdroid android apps too.
The projects that are actually only AI assisted don’t get that many complaints on most communities I see. It is the slop that people push as legitimate that gets people riled up.


Also the controlling party in Italy is practically neo-fascist and Austria is also sliding there again with their Nazi party like AfD in Germany and VB in Belgium getting a huge portion of votes.
I don’t see how it is that much better than a US non- profit.
I think the core thing is that you have to know of a bunch of different CA options and switch when necessary or on a compromised CA.


Also, looking towards the future or for people that can find one at a decent price:
Not a raspberry pi but an Orange Pi Plus or Pro. They have a rockchip that has AV1 decoding (along with standard others) which is probably going to be the future (it is for corpo streaming already).
Put plasma big screen on it and get a Bluetooth wechip remote or something


There is no way this isn’t just an AI bot


I guess there is no rule about having the AI writing the AI disclosure because I guess 4 sentences has to be extended into 8 paragraphs.
But what if the AI disclosure lies? 😂 for example, PRs are reviewed by Claude code and merged. That doesn’t sound like “every commit is reviewed by the maintainer”


This seems like the perfect scam for Lemmy for setting up a fresh IP for launching a cyber attack/scam with a real residential IP. Helping a community member while circumventing shitty corporations. (Not saying OP is, just that it would be a decent scam)
This requires either an insane amount of trust in a random person or gullibility. Plugging a random mailed device into your network, even behind a VLAN.
I definitely would never do this if I didn’t have absolute full and complete access to every part of the pi’s software, or preferably set it up myself (optionally in collaboration with the requester) and the requester only had read-only access to it.


I am already trying to understand all of it first myself. In the middle of a full-strip renovation; I don’t think I will have time to spend on it 😂


Please dear god let STM or Nordic someone hire some technical writers for Zephyr, not just Nordic documentation.
The data structures, macro basics, and core concepts are OK, but for example ADCs have literally 0 documentation at all, documentation about DMAs is barebones at absolute best, non-led PWMs is not good, etc…
Pretty much every single peripheral needs 5x-100x the documentation it currently has lol.


Some people are fine with down time/inconvenience in exchange for security.
I have my boot drive on a secured USB and LUKS keyfile with the rest of the partitions on an encrypted SSD and data on encrypted HDDs.
In a smash and grab (or fascist government gestapo smash and grab), the server is pretty impossible to steal information from (inject illegal content to in order to fabricate evidence) without the USB and they can’t simply inject boot malware either. A network device is almost always findable either by cables or WiFi broadcast analyzing.
Ah so it isn’t the kopia server receiving the backup, it is the client making a server backup to a dedicated drive? Not connecting to a different computer’s kopia instance
Obsidian doesn’t really have self-hosting as far as I know, even though it is an electron app so it would be trivial for them to do, but it would cut into their revenue stream of obsidian sync. One of the downsides to closed source. The only things related are static website “wiki” generation from the markdown file collection.
Quartz https://github.com/jackyzha0/quartz
Or people use it for blogging like
Hugo and Jekyll
Syncthing (or nextcloud/Opencloud) can be used to sync obsidian vaults across devices though.
I would suggest opencloud over nextcloud as it is quite a bit faster and not just a mass of extensions and tends to break a lot less than nextcloud. I stopped using nextcloud because of the amount of times it completely broke on an upgrade while not even having breaking changes (I’m glad I didn’t have end to end encryption)