sorz

They do test continuous (1-hour) gaming performance in controlled environment regularly. Thermal throttling can be easily spotted on the frame rate vs time graph.

I just come back from a furry con hosted in hotel in China, and learn how every furry con like this completely break hotel's elevators:

- rooms are all booked

- young furries usually share one room with 4 or even up to 6 friends

- more guest room-lobby-ballroom traffic

- far more inter-guest-floor traffic than their usual tourists

- one full-suit furry take 1.5-2 human's space

Their elevators are not designed for this scenario and got overloaded all the day. Queueing for tens of minutes is not uncommon.

So gound stations need keep measuring their own positions due to continental drift! Never thought that before. Thanks.

I found one taken with Z9: https://images.nasa.gov/details/art002e009301

There are some very bright noise pixels on the dark area, which is different from the noise in similar photos taken with D5 (much darker and uniform).

I feels wrong to not mention IPv6 in 2026.

- net.ipv6.conf.all.forwarding=1

- nftables is default to `ip` family which only applies to IPv4. Setting it to `inet` will allow rules to apply to both IPv4 & 6; or `ip6` for IPv6 only. You can skip NAT rules, usually.

- dnsmasq: in addition to DNS and DHCP, turns on router advertisement with SLAAC. Some devices can get IPv6 address from stateful DHCPv6 server, others (e.g. Android) only work with SLAAC.

Lots of issues follow the pattern "ANSI escape code inside untrusted text". It feels like XSS but for terminal.

Random thoughts: one can get user's ssh public keys from GitHub on the fly (from `https://github.com/<username>.keys`), so that it requires a valid GitHub account to use this service, without (extra) auth process.

Better be "three months of the country's GDP".

The time at which scalpers buy tickets becomes irrelevant. They can resell tickets at arbitrary-high-price + margin because buyer know they will receive refund for the exceed amount, thus pay only (final price + scalper margin) for the ticket eventually.

Bonus: buy ticket from scalpers after price settled, you will get a determined price, no more guessing and find inner peace.

The problem for scalpers is that if they buy too many tickets, the final price may become too high to be attractive for real buyers.

In the mug-scrubbing video, the person clearly pretends to wash the cup but does not seem to want to get their hands wet anyway. I'm curious as to when models can figure out that subtle thing.

Did you try scancode map registry[1]? I use it to swap Ctrl/CapsLock and it works well even with RDP softwares and run-as-admin apps which PowerToys couldn't handle.

[1] https://superuser.com/questions/550679/where-can-i-find-wind...

The whole post is on base on the fact that "TW" in their URL. But I think this is really just a internal implementation detail of Google Translate, which is not intended to display to users (e.g. complete hidden on their app). Everywhere on their UI shows "Chinese (Traditional)". That is what they try to communicate with users.

Sure, zh-TW is somewhat misleading. But they nor say that parameter is a ISO 639 or RFC 5646 conformed.

They said the paper is still working in progress and will improve it.

https://twitter.com/AiEleuther/status/1660811180901019648

Just watched a talk[0] about natural language understanding research in post-GPT-3 era. Old issues may has been solved, while new topics are coming to this area (quoted from the slides):

- Retrieval augmented in-context learning

- Better benchmarks

- Last mile for productive application

- Faithful, human-interoperable explanations

[0] https://www.youtube.com/watch?v=-lnHHWRCDGk

Is NLP a solved problem now?

> then the amount of information we need to make billboards and signs make sense.

Subsequently, this applies to posters, letters, newspapers, and other types of text-heavy images, ultimately reducing the language modeling problem to an image generation problem.

OpenAI did exact the test for GPT-4. The raw, non-fine-tune GPT-4 is quite good at predicting confidence level ("highly calibrated" by their words). But the RLHF fine-tuning process seems ruin its calibration. Figure 8 on page 12 of GPT-4 Technical Report shows this dramatic changes before & after fine-tuning.

Is it possible that the profile is over-fitting to the benchmark tests?

Does it work for WPA3-Personal network?

Want to know when Firefox will support Web Serial API, but turn out they won't since they consider it harmful: https://mozilla.github.io/standards-positions/#webserial

A: Look this, the perfect programming language I invented is the best.

B: (tick the checklist and sent it back)

> Won't I have to reload my browser if I change a file?

> Yep! Welcome to the joys of Javascript.

Although it (re-)sends file for every request, the file content must be identical. So the relay server may just cache the file.

Once you do that, it becomes more like a traditional static file hosting w/ on-demand file uploading via browser.

It even works on my latest Firefox beta on Windows 10 (on this HN page). Hmm...They may not the same thing.

dnsmasq also has out-of-box support for adding DNS replies to ipset: the --ipset option.

> Windows Firewall worked on WSL Pico processes for example.

In WSL1 they share the same IP addresses and TCP/UDP port space, while WSL2 has an separated IP address. I suppose there is some NAT to make the network working in WSL2.

At the end of Q&A part they mentioned that sharing localhost, IP addresses, and port number space (which is a WSL1 feature) may be done in future, but they have no roadmap for it right now.

No longer since TLS 1.3

I saw that strongSwan had already supported two post-quantum key exchange algorithms (NTRU and NewHope) for IPSec IKEv2. Good.

https://wiki.strongswan.org/projects/strongswan/wiki/IKEv2Ci...

If you want remove all China CAs, it may be helpful. https://github.com/chengr28/RevokeChinaCerts