How does it go with your wife when opinions are opposed and the matter is important (such as kids)?

You're behind by 1.5 years on that thought. They certainly can.

If you have latency detection already why not pause H.264 frames, then when ack comes just force a key frame and resume (perhaps with adjusted target bitrate)?

That would require that they understand the protocol stack they're using to send H.264 frames

Yeah, monitor the send queue length and reduce bit rate accordingly.

This picture is wrong because deaths themselves are not as interesting as estimated years of live lost.

For cancer and heart disease that multiplier would be around 5x (e.g. mostly older people die from cancer, but if they did not, in ~5 years (optimistically) they'd probably die from something else anyway). But for overdoses that would be around 50x. So one needs to scale overdoses at least 10x.

I did the same with recent stuff and so far gpt-oss-120b on high was the best with gpt-oss-20b on high close second.

Are you sure about OpenAI? I thought they actually do retain your agent chats (training I am less concerned about personally).

Anthropic has an option to opt out of training and delete the chats from their cloud in 30 days.

I was only talking about training so you're probably right about retention - I care more about training.

I bet on km/h vs mph mistake.

This makes me think: are there any "pair-programming" vibecoding tools that would use two different models and have them check each other?

Both your examples require Internet access so there's no reason not to use cloud-hosted model which would work magnitudes better.

Would have to be ChatGPT these days.