They wouldn't even need to do this uniformly, quantized versions of the model could be routed only a subset of the requests. They could do this to nerf the old model, or more likely just to give themselves more hardware to run the new one on by handling more requests on less hardware. Or to handle increased request volume as traffic ramps up faster than hardware can be provisioned.

Playing with local models at various quants, the degradation can be hard to spot. Sometimes it's only noticeable in aggregate. And even then, you never really know if you just got unlucky with a bad response due to RNG.

I've had Opus 4.6 fall into some weirdly incoherent loops that I rarely see from even Sonnet, that felt like the kind of thing I got frequently with Qwen3.5 9B on local. And the above applies... Was that just bad RNG? Or was my request to Opus routed to some lower quality variant? There's no great way for me to tell for any given request, nor any way to guarantee Anthropic _didn't_ do that.

To make 4-bit fit on one card with reasonable (100k+) context needs a bit more care though. And tuning can be highly specific to your machine, gpu and use-case. But I use a headless server, offload multi-modal to CPU, use fit-target to reduce wasted memory and use q8_0 kv since the 4090 performs well with it... In addition to most of the same config as the author elsewhere. I get 50-60tps generation with a power limit of 275W (450W is default), more than enough to offer a roughly an Opus-speed feedback loop.

I haven't seen many of the issues with looping the author mentions. But I did with Qwen3.5 and in particular other 4-bit quants in the past. But the difference is probably a mix of the improvements above, as well as habits changing to avoid cases where models will loop. For what I'm doing, it seems like I loop Qwen3.6 on the same kind of prompts I'll make Haiku or Sonnet loop on (the latter hide some of their existential loops behind "thinking"). Usually it's cause I was too vague about some aspect of what I'm wanting them to do or I forgot to include some context that smaller models just don't have access to in their smaller knowledge base. But at least for what I'm doing (Rust, React, kubernetes) it's not been a notable problem at all with the latest iteration of this whole stack. And knowledge of standard libraries and default k8s resource kinds has been almost flawless.

There's still plenty of more complex stuff where I'll choose to jump straight to Claude or GLM-5.2, but if it's not worth that jump I've stopped paying for the middle ground as it's usually not much better than just one more iteration through qwen.

All this to say, if you have a 3090/4090, feel free to give the same setup a go. It's come a long way in recent weeks.

In that sense, seeing if the latency impact of postgres is tolerable is pretty reasonable. You may be able to get away with postgres putting things on disk (yes, redis can too), and only paying the overhead cost of allocating sufficient excess RAM to one pod rather than two.

But if making tradeoffs like that, for a low-traffic service in a small homelab, I do wonder if you even need a remote cache. It's always worth considering if you can just have the web server keep it's own cache in-memory or even on-disk. If using go like in the article, you'd likely only need a map and a mutex. That'd be an order of magnitude faster, and be even less to manage... Of course it's not persistent, but then neither was Redis (excl. across web server restarts).

So saying it's not waves is dangerous, and saying it's not a sea level rise is dangerous. It's not useful to try and delineate between a tsunami being one of the two when it's in reality an event that consists of both.

(Ignoring that a sea level rise and a long-wavelength wave are the same thing)

The news mentioned a previous similar event where the largest wave was 4 hours later.

Waves can get bigger due to earthquakes not being instantaneous or necessarily a single movement, due to amplification by geography, by reflections, by aftershocks, and many other things. The news is suggesting waves lasted about a day for a previous event in a similar area.

Being able to chat casually with low latency, correct yourself, switch languages mid-sentence, incorporate context throughout a back-and-forth conversation etc. turns talking to these kinds of systems from a painful chore into something that can actually add value.

This is a bit of a stretch. There are cars everywhere you go in Tokyo, it's pretty well set up for driving given the size and population. That said, I've lived here without a license for years, and rarely had the need to hop into any cars. Speed limits are generally low, and lights are everywhere, which often makes the train or sometimes even a bike a faster option than a car or a bus.

It's only when you have multiple connections in your route or when you get well outside the city that you start to see consistent benefit from a car.

One reason it's so easy to get by without a car in most of Tokyo is that the shops and attractions are distributed well. Zoning means there are tiny shops everywhere, and the bigger shops are present at many of the train stations in the city. You are pretty much always within walking distance of everything essential. One more reason is that the postal system is excellent. If you order something from Amazon, you'll receive it about 24 hours later in most parts of Tokyo. Who needs a car when you pick up everything essential on a 10 minute walk from your house, and everything you want will be at your door tomorrow?

Also, there are plenty of terraces to have a nice coffee on. The difficult thing in Tokyo is you need to search in advance for what you want. It's so crammed full of options, that you might never realize all the interesting things you're walking past. It's a very three-dimensional city, there's stuff hidden everywhere.

  github.com ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAQEAq2A7hRGmdnm9tUDbO9IDSwBK6TbQa+PXYPCPy6rbTrTtw7PHkccKrpp0yVhp5HdEIcKr6pLlVDBfOLX9QUsyCOV0wzfjIJNlGEYsdlLJizHhbn2mUjvSAHQqZETYP81eFzLQNnPHt4EVVUh7VfDESU84KezmD5QlWpXLmvU31/yMf+Se8xhHTvKSCZIFImWwoG6mbUoWf9nzpIoaSjB+weqqUUmpaaasXVal72J+UX2B+2RPW3RcT0eOzQgqlJL3RKrTJvdsjE3JEAvGq3lGHSZXy28G3skua2SmVi/w4yCE6gbODqnTWlg7+wC604ydGXA8VJiS5ap43JXiUFFAaQ==

[1] https://github.blog/changelog/2022-01-18-githubs-ssh-host-ke...

[1] https://github.com/kprotty/zap/blob/blog/benchmarks/rust/ray...

[1] https://github.com/Chia-Network/chiapos

Living in Japan now, I had a few months with a padded mat + quilt on the floor as is tradition (and a damn cheap one), but upgraded to a mattress on the floor because the floor was too cold in winter as you mentioned.

There's as much to be gained from taking stuff away that isn't useful, as there is from adding useful stuff to your life.

There are so many better things to spend time on than the mundane parts of life.

The index fund investing with scheduled transfers is exactly what I meant by automated financial services. I probably micro-manage it a little too much right now for no real benefit.

I see a lot of comments that seem to see all the things you miss out on in this situation. But in my mind, it frees up a lot of mental effort, time and stress. If I ever get bored I can go to a restaurant and eat something wild and it will be all the more exciting given I don't optimize for excitement or luxury in my everyday steady-state.

When Soylent came out I was super excited about this idea. Don't think about three meals a day that you normally fuss over, and instead have two predictable, quick meals and optimize to make the third one amazing. Soylent was OK, and DIY soylent offered some hope too. The third meal WAS always amazing, in a relative sense, and tasted better somehow than when I had the same thing before this diet. Unfortunately liquid diets are just not satisfying to me and so frozen meals won out.

I'd love to find other areas of my life that can be similarly optimized. I have hope for bill management services to take the annoyance out of juggling payments etc., and roboinvestors or similar automated financial services. Doing these things manually offers no excitement and no added value beyond the transitively provided service so I don't think they should take up my life.

The amount of time wasted across the whole human population on things like preparing meals, choosing outfits and managing everyday responsibilities must be huge and that is all time that could be spent doing other exciting or valuable things.