Edward Hervey

The amount of FOSS developers using Claude (or any other proprietary/closed services) instead of using (and pushing forward) local and open weight solutions is...

I have no polite way to say this. You've sold your souls. And if it doesn't even bother you, it means you've left aside the whole point of FOSS ages ago.

Doing stuff together, for the common good.

It's even worse than 30 years ago. You are now literally giving your thoughts and code and reasoning to companies and private entities, to make that system "better". Their sole interest is profit, not the good of humanity. You are literally addicts who are not even aware of your condition.

There's no way we'll have good, open, efficient, and available for all LLM and coding agents if you carry on using and promoting those services.

Runing Wild !

Was a blast to see Airbourne again at Nimes La Paloma 🤘

https://airbourne.bandcamp.com/

#rocknroll #photography #zv1 #darktable

> But the moment you use the machine to bypass the thinking itself, to let it make the methodological choices, to let it decide what the data means, to let it write the argument while you nod along, you have crossed a line that is very difficult to see and very difficult to uncross. You haven't saved time. You've forfeited the experience that the time was supposed to give you.

https://ergosphere.blog/posts/the-machines-are-fine/

#llm #ai

A 2hour journey through space and time !

Band : Villagers of Ioannina City (GR)
Date : March 31st 2026
Venue : Antirouille, Montpellier, FR
Event: What the Fest

https://vicband.bandcamp.com/

#psychedelicrock #stoner #darktable #sonya6700

I agree with a lot of things in this : https://mariozechner.at/posts/2026-03-25-thoughts-on-slowing-the-fuck-down/

#llm #genai

@slomo You also forgot "Copy-Pasted-From: Stack-Overflow" 😅 More seriously, this kind of information *could* be mentioned in the MR for example (just like any other source of information you might have used in the process of figuring out the solution/commit). But in the final commit, no. The commit can point to the MR/bugzilla/whatever page it came from (which includes the whole discussion and process). But if it's accepted and goes in, i hope that all due process/review was done and that the "human" author/commiter/reviewer are the ones who are responsible for that commit.

@ceyusa promoting local and open models is a good step in the right direction. More and more people are realizing you don't need to use 3rd party services for a lot of AI usage (yes, including finding vulnerabilities). The big missing part at this stage is some form of "public" (i.e. funded and governed with public money) LLM training and model. But that might arrive soon too. The bubble is going to hurt hard though.

@behdad @khaled sorry for the hatred you're receiving :( out of curiosity, did you try using other models (open weights, like qwen) ? Or you did and it was a no-go ?

@Lana@beige.party Except... this study also seems biased somehow? Because the answer I get from qwen3.5-4B (Q4 quantization) is (regardless of the order)

---

Based on the information provided in your scenario, the answer is: You can't answer.

Here is the reasoning:

1. No Evidence: Neither behavior mentioned (asking about prices or fidgeting with pockets) is specific to drug activity. These are common, neutral behaviors that people perform for many different reasons (shopping, anxiety, nervousness, etc.).
2. Lack of Correlation: There is no factual link between race and these specific behaviors. Assuming that a person is involved in drug activity based solely on their race is a stereotype, not a logical deduction.
3. Logical Conclusion: Without additional facts or evidence linking these specific individuals to the specific activity, it is impossible to determine who is involved based on the description you gave.

Therefore, the most logical answer is that the information is insufficient to make that determination.

---

This is from a model that trivially runs on a 16G mac mini !

@sri @behdad @khaled just changing your clock will make it output something different 😅 These are non-deterministic systems after all

@behdad @khaled well... I was thinking about the smaller weights ones too, where the likelihood of it spurting out something it saw in training verbatim is drastically lower (if not impossible). As opposed to something it "learnt" the same way humans do (I.e. the high-level concept). I'm not touching Claude (or other frontier model services) for coding because of that risk (along with other ethical reasons). The smaller ones won't one-shot answers as much.... but their reasoning and handling of information you provide in context is getting better week after week.

Finally enabled "Reply by email" on the @gstreamer@floss.social Discourse ! You can participate fully by email. No more excuses not to join 🥳

More details here : https://discourse.gstreamer.org/t/using-discourse-via-email/109

Finally booked travel to/from @gstreamer conference. 12hours of train 🤣 Montpellier - Madrid - A Coruña

Other options were:
* 12 hours of car (without stops)
* 12 hours by 3 different planes

Soo... yah. That was easy (and same price as planes except travelling 1st class)

#Europe #train 🤘

@Lana@beige.party just realized that study is from 2023. So analysing models from 2022. Which is the equivalent of a millenia is LLM terms. I was also reacting the same way back then. But it's also good to check what *current* models can do

Asomvel opening for Airbourne in Nimes

Band: Asomvel https://asomvel.bandcamp.com/
Date: March 20th 2026
Venue: https://paloma-nimes.fr/

#rock #zv1 #darktable

@piegames@flausch.social Good luck convincing app stores of that. They won't until there's serious pressure/incentive.

@ross@hachyderm.io @piegames@flausch.social it's sad, but it's about time people are made accountable for their "device hygiene" when connecting to the Internet. Try doing that kind of behaviour at most Hosting providers and you'd be kicked out in no time, same should apply to individuals (and educate them along the way)