The summary seems wrong, the Qwen-2.5-max blogpost compares against v3, not r1. This Qwen model doesn't appear to be trained to reason, unless I missed something. It's just a traditional LLM base model.

No, only that people can't read, which is nothing new. This is the literal quote from the deepseek v3 paper: "Note that the aforementioned costs include only the official training of DeepSeek-V3, excluding the costs associated with prior research and ablation experiments on architectures, algorithms, or data."

In other words, this is only what deepseek has been saying from the very beginning in its public tech report.

Try that on the Earth sometime. :-) I have hunted my share of radiosondes and this is extremely tricky in practice unless you have a direct line of sight. The signal gets obstructed by terrain and reflects from surfaces.

The main point is to save time on porting, from OS-specific APIs to maintaining constantly breaking cross-compiling chains for OSes you never actually use personally.

Exactly, and N26 or Revolut are getting pretty big in Europe. They are app-only with literally no physical branches. (Revolut isn't quite a bank just a credit card though that's often enough, but N26 is a real one.) It turns out that banking can be quite simple, a breeze with a well-designed app, and completely remote.

Huge scale has good aspects as well. It becomes standardized, making it easy and efficient to develop solutions (plus troubleshoot etc.). It enables things like massive elasticity.

The control imbalance is a downside. How much it's worth any extra regulation? I'm not sure myself.

China *already* proved you wrong. Even the beggars in the streets wear alipay QR codes. For daily use, cash already effectively went away in China *right now*, not in the future. No reason this is not going to happen in the rest of the world, Africa might be next. No opinion when it will happen in the US in particular, but who cares.

Are they missing few decimal places with the emails? Those 92000 emails I have in just one of my inboxes are for just a couple of last few years. Do I have a problem? :)

> Now, miners can apply all the hacks they want (SegWit and co.) but if the system wasn't made to scale, no amount of patching will make it scale.

Unclear why this would be the case. The Lightning Network should compress many transactions to a single one per a long time period.

The slowdown is on syscalls. So it depends on your workload. For example, `du -s` is reportedly slowed down really by tens of percents.

Lightning Network (http://lightning.network/) will allow Bitcoin to scale. Maybe not to VISA-level yet, but at least further several order of magnitudes.

The lightning network allows to keep most transactions off-chain - for example all your daily payments can go through a LN operator, and the only thing that needs to be stored in the blockchain are channel balance settlements - say once per month you settle the balance with a single transaction. It recently left beta, with the protocol finalized and three interoperable implementations. Now, UX and client support is needed, and adoption by major Bitcoin operators. But that'll happen.

Bitcoin adoption just went a bit ahead of the software ecosystem, but don't discard Bitcoin because of that just yet.

Meanwhile, Public Backdoor in Many Chinese Phones Sent Data To US.

...yet, there are professional chess (and Go, shogi) players. It's just smaller economics than football, but that's okay.

Which is all right since the cube root of 1e6 is 1e2. The previous best algorithm would require 1e4 pages.

Siri is a complete stack of text reco engines, intent recognition tools, and backends. There are many initiatives like Sirius, Mycroft and YodaQA, and each does something slightly different - either focusing on the speech reco infrastructure, or just answering factoid questions...