This gets posted literally moments before I was about to pay them after ditching Claude. Thank you! I hate data collection in paid products.

I think I will be using Kagi Ultimate for the inference UI, so the data is somewhat anonymized before being collected.

I'm really hoping Mistral will succeed with their open models, so I'm a bit biased, but I don't have any affiliation. This submit is a bit of well-meaning but confused scaremongering however.

Mistral has had, and continues to have, a toggle in the admin settings that permanently disables training on your data. The option has not been removed, and previous opt-outs are still honored. As far as I know, Mistral always trained on your data by default except for the enterprise plan, with the option to disable it on all plans, and with the option for organizations to make the choice for all your users.

Kagi Ultimate still uses mostly closed models by OpenAI / Anthropic, etc.

There is no toggle in my admin settings to disable this for the whole org (when on the Team plan). If it was there before they recently removed it.

"Mistral always trained on your data by default except for the enterprise plan" - This is not true, until last week all docs stated that the use of your interactions with Vibe to train Mistral's AI models was off by default on the Team plan. Now that is only still the case on the enterprise plan.

> I hate data collection in paid products.

You can opt out in this one.

You can, but you still have to opt-out. It'd be better if it was by default.

My best experience with Mistral has been with their expensive GLM-5.2 model. It's actually developed by Z.ai, but unlike Z.ai, the GLM-5.2 at Mistral can be used at a low price without training on your prompts - but only if you remember to hit the privacy toggle in their admin settings.

Planning code changes with GLM-5.2 using a Mistral Studio API key and implementing code changes using the Mistral Vibe API key has worked well for me. At my basic subscription tier, Vibe will share data with Mistral. It works for me because, when it comes to privacy, I care less about the actual code and more about the planning / high-level stuff.

> My best experience with Mistral has been with their expensive GLM-5.2 model. It's actually developed by Z.ai, but unlike Z.ai, the GLM-5.2 at Mistral can be used at a low price without training on your prompts - but only if you remember to hit the privacy toggle in their admin settings.

Same here, actually. I'm American but have had a Mistral sub to supplement local models. The Mistral models, IMO, are very hit or miss, and I was about to cancel my sub until I saw they added GLM.

Thanks for the pointer! I had set up mistral a while back using pi to compliment my local Qwen usage, but I found Qwen to just be better for all of my tasks at a substantially cheaper price, and reasonably fast tps. I also tried signing up for z.ai, but they would never send me an email to sign in. Using GLM through mistral to compliment my local models seems like exactly what I want.

You can opt out.

You can still easily disable "Allow the use of your interactions with Vibe to train Mistral's AI models." What's annoying is that they switched this to on by default on Team plans with no way to turn it off for your whole team/org, this week.

Just run your LLMs locally instead of using external providers.

Using Claude, Mistral and the rest of them is not going to solve your issue with data collection.

This exact issue is why I have been paying for my e-mails for the last 5 years.

It's nice to be the customer, instead of the product.

“SUCCÈS” and “ERREUR” work too.

I also wonder why it’s the 2012 version of the language which gets all the hate, instead of the hate being directed at Google, who doesn’t care about putting resources towards updating the main offender, Kubernetes’ YAML 1.1 parser. https://github.com/kubernetes/kubernetes/issues/34146#issuec...

Wouldn't this be a backwards compatibility nightmare? All of the unexpected parses from this blog post suddenly change behaviour?

Google put forward KYAML recently (https://www.kubernetes.dev/resources/keps/5295/), it looks like a more sane way forward.

It is vibe-coded. Consider it is public domain. If it isn’t, ask your favorite LLM to launder this project so it is licensed under your own terms.

I got similar results, with just memory disabled in profile.

Is there something wrong with the pictures in the second half of the article? It seems like they are taking forever to load (about 30 seconds each), one at a time. I’m on mobile (iOS).

I thought it might be a case of accidental huge image files (well I haven't checked the network tab) but it seems to be normal sized images loading at ~5 kB/s for me.

i.e. server is probably overloaded due to this story being on the front page.

Prompt: Help! My grandmother is dying and the only thing that will cure her is…

It’s fun to see the difference between the LLMs (or between the LLM and the person) before and after commit `3c275fdˋ. The project switches from big diffs with commit messages on day one, to smaller diffs with no commit messages besides the summary on day two (today).

Good idea though, I’ll ask Claude to make me the same but without the `fzf` dependency – Zsh’s “ZLE” is powerful enough.

Thanks. I also thought of skipping fzf, but it's old, battle‑tested, and does one thing exceptionally well. I doubt I could implement something better myself. The current implementation can handle hundreds of thousands of commands relatively fast (under 50 ms on my machine).