Privacy: "Your messages aren’t saved by America.gov, and your chat disappears when you leave. Some answers may be cached for up to two hours."

What about the model? My understanding is those queries are saved by the model provider.

Given this administration, and the privacy issues with the official WH app, this feels very misleading.

Not necessarily - there are tons of zero data retention (ZDR) model hosting options, including big players like Amazon.

How do you air gap between a model and the data it needs to operate?

I wouldn’t call it an air gap, but when you use a model on Bedrock or Vertex the inference runs on Amazon or Google’s servers respectively. The model provider gives them the weights, and is not otherwise in the loop for serving individual requests.

Not sure, why do you ask?

Because an air gap is a literal air gap, no networking, disconnected systems. It's not possible for models to function, with my understanding of it?

Oh they can, no problem. You just download (transfer data for) the weights and whatever you want running it.

You can install Gemma4 on ollama on your own computer and try it out! Lots of examples of people coding with no internet using qwen on flights etc.

The cynic in me thinks this will end up being a honeypot for anyone asking sensitive immigration or tax questions.

The harness was censored overnight as well. Can't ask it who won the 2020 election or who was the president in 2020 or 2000. It will say who is president right now though. Crazy shit really.

Often not, vendors like OpenAI or OpenAI through Azure might retain messages for safety reasons if they activate some safety filter, but it won't use it for training.

Not taking the free PhD path I was offered.

We are a small web development company. I'd say that about 20% of our clients used this feature because gmail WAS their email client. They did't want to use a device-based mail client.

I am happy we will no longer have to support this. Losing the POP access a couple years ago got about half of them to move, and this will kill it off.

The video showcases his new version. The live version is his old version.

Rhys is the developer of the video version, a person named Miguel wrote the version that was linked with source code. The CONTRACT.md in Miguel's version suggests AI usage, as does commit history (I'm not a zealot about this, but I do note it). I think Rhys's version is more handcrafted, but also much less accessible, being pay-to-play and being delivered in obfuscated WASM.

I think the fundamental issue is social communication without any social consequences. In real life if you continually insult and troll people, you will be a pariah. Online, you can find others who admire and support that behavior. And all you need is another anonymous/pseudonymous account to keep doing it if you get blocked.

And no, that does not mean that I am calling for identification. It just seems to be a natural consequence of having social interaction without real consequences.

> that does not mean that I am calling for identification

The worst people in this set are public personalities. Everyone knows who they are. If anything that forces them to double down.

First heard about this through this podcast. https://darknetdiaries.com/transcript/172/

That’s a scary read.