Go up a level in the URL, or click on the logo in the upper-left on the "try" page the post links to.
1,413 karma · joined November 12, 2011
(Even the NY Times gift link requires a login.)
Fun fact, though: You have to replace them every couple of years (or sooner) because no pigment/dye is stable, especially if exposed to bright light.
Having a test shot with one of these in it saves you untold amounts of pain later.
The "ethical" employees will think they'll solve the problem later. The unethical ones won't be encumbered by such thoughts in the first place.
https://en.wikipedia.org/wiki/Hidden_sector
The scary thought is that, were it true that there is voluminous richly-interacting dark matter, we would be the actual dark sector.
For comparison's sake, Qwen 3.6 35B A3B MLX, same prompt, same machine, thought for 1.59 seconds, used 2,398 tokens, and hit 80.83 tok/s. It actually output just a straight SVG file of an owl (not a bad owl either!), whereas Qwen 3.8 27B spent a zillion times longer gold-plating everything and output HTML+JS+SVG with interactive animation, sound, etc., etc.
The MoE models are stupendously faster.
Note that my mini is the bigger Pro model, so it will be faster than the mini quoted in the article.
Mac mini M4 Pro (cores: 10P/4E/20G) 64GB Tahoe 26.6 LM Studio 0.4.20+1
Qwen3.6-35B-A3B-MLX-4bit: 78.81 tok/s, TTFT 0.93s (but note Qwen 35B is really chattery and outputs 3,451 words of thinking for 65s first)
Qwen3.6-27B-MLX-4bit: 14.65 tok/s, TTFT 0.76s (Qwen 27B output 3,027 words of thinking for 361s first, spinning up the fans)
gemma-4-26B-A4B-it-QAT-MLX-4bit: 64.93 tok/s, TTFT 0.44s (Gemma 26B is much more on-task, thinking with 591 words for 15.62s first)
gemma-4-31B-it-QAT-GGUF Q4_0: 12.25 tok/s, TTFT 1.65s (Gemma 31B thought with 404 words for 51.55s)
I'd love to buy an off-the-shelf WiFi+NTP clock, but everything I've seen is very expensive, intended for commercial timing applications or synchronizing clocks across a campus.
https://news.ycombinator.com/item?id=48910545
Perhaps they got tired of LM Studio, etc., not being able to run their models properly.
https://community.letsencrypt.org/t/dns-persist-01-deploymen...
I wonder if the interim version has been rolled out to some CAs.
https://github.com/acme-dns/acme-dns
CAUTION, though, the last time I downloaded a binary release, ClamAV triggered on it, so I kept my old version which worked. I was using the 1.0 series (without any problems!), and now it seems the project has picked up development again with a 2.0 series.
This is not the language of an elitist.
If anything, it sounds like someone defending Liminalism's inclusion in the contemporary canon from arrogant elitists.
This could also be an unofficial-official way for Apple to "leak" that yes, they do this--which is on brand for how Apple handles "rumors" etc.
They cannot envision the scenario where their AI-powered robots turn on them, or at the very least are used against them (and then inevitably turn on everyone).
Edit: For comparison with the other poster, same setup as above, but with Gemma 4 31B Instruct 8bit MLX (not sure if exactly the same model): time to first token 4.62s, 7.20 tokens/s; with a different prompt, 1.17s and 7.24 tokens/s.
That said, I agree with you!
First, they can't attack a WiFi access point for which they do not know any password(s). Thus your multi-SSID access point with multiple passwords is "safe" from this particular attack.
However, second, they can attack an access point for which they know any password, gaining access to clients on the other SSIDs. This means your security is now effectively only the security of your worst SSID's password. It also may defeat your purpose in having multiple SSIDs/passwords in the first place.
Granted pronouncing the name is ambiguous, Wes-cam or We-scam. But they're known well enough in the industry at this point that it's not a problem for them.