SloopJon

I somehow expected more density for a custom arrangement like this. One eviscerated mini per sled, ten sleds per rackmount chassis, six chassis per rack: sixty minis in 42-48U.

Sonnet's RackMac mini[1] puts two unmolested minis in a short 1U enclosure: sixty minis in 30U. A web search turns up the Apple Mac Mini Relay Rack Shelf[2] from RackSolutions, which has four minis in a full 1U enclosure.

I'm willing to believe that there are cooling and airflow advantages to the sleds, although those two small fans at the back of the sled don't look too exciting. The biggest advantage seems to be cable management and the ease of sliding out a single mini at a time.

One comment on the video points to another video[3] about MacStadium, which mounts minis sideways on what look like library shelves.

[1] https://www.sonnettech.com/product/rackmacmini.html

[2] https://www.racksolutions.com/mac-mini-shelf.html

[3] https://youtu.be/0b46E4mp_V8

Depending on what you want to do on what O/S, batteries may or may not be included. I use Moonlight on a Mac to control a Windows 10 PC running Sunshine--lower latency than Remote Desktop (which I don't have anyway in the Home edition), and nicer looking than Parsec.

When you connect a patch cable directly (no crossover cable needed in the 21st century), you'll likely find that each system has a self-assigned IP in the 169.254 network.

I've been playing with Thunderbolt networking over the past week with mixed results. I can get 16Gbps between a couple of Macs. Between a Mac and a PC running Windows 10 I get similar speeds in one direction, but less than 1Gbps in the other direction.

In terms of scaling this to multiple hosts as the author does, I've read that it is possible to daisy chain, or even use a hub, but it doesn't strike me as the most reliable way to build a network. For an ad hoc connection, though (like null modem cables of yore), it's a great option.

This may be the longest Stack Overflow question I've ever seen, with a code block, three graphs, and a list of links to related questions, comprising four and a half screenfuls on my display. (The accepted answer is five and a half screenfuls.)

The estimated value of $20K seems low, unless it's in poor condition and/or there are many other copies around. Google says the same auction house sold one for $100K in 2012:

https://www.reuters.com/article/idUSBRE8AT04H/

Although this post doesn't establish the premise sufficiently to comment on the data or its schema, I would venture that much of the time what you want is a GROUP BY query on columns of the table, rather than a column of pointers.

What distinguishes a Foo from a Bar? Are 1, 2, and 3 (or I, II, etc.) actual data values, or simply row numbers arising from a sort on something else?

If you're writing the back end for OmniOutliner, and you must use a relational database, then maybe you'll use tricks like these. Otherwise, I would see whether you can come up with a schema that better models your data.

See "Chrome Sync privacy is still very bad"[1] and its predecessor article[2] about Chrome and Firefox for an evaluation of the privacy and security of browser syncing.

[1] https://palant.info/2023/08/29/chrome-sync-privacy-is-still-...

[2] https://palant.info/2018/03/13/can-chrome-sync-or-firefox-sy...

There is an FAQ for the mp-units project on which this proposal is based about why they chose not to use user-defined literals:

https://mpusz.github.io/mp-units/latest/getting_started/faq/

If I understand correctly, though, the unit expressions that manifold implements aren't limited to literals, is that correct?

Note that this is heavily dependent on C++20. I'm not even caught up on C++17, so I was thrown off by statements like this:

    quantity q1 = 42 * J;

where quantity is a class template:

    template<Reference auto R,
        RepresentationOf<get_quantity_spec(R).character> Rep = double>
    class quantity;

I didn't know you could use auto in a template parameter. I gather that Reference is a concept, something else I've yet to explore. Just when I thought I'd gotten the hang of "modern" C++.

I believe that "^" in this context signifies Caps Lock as its own modifier, distinct from Control. If you've already remapped Caps Lock to Control (or Esc if that's your thing), this scheme won't work without some modification.

There's nothing scary about being downvoted on HN, but if karma is your goal, I wouldn't start by labeling a simple (and frankly 100% expected) fork a holy war.

I'm not sure I understand your concerns in the previous comment. What would cause you to constantly update your YAML, and why are you just now worried about a rogue dev? Also, who doesn't trust whom?

I see three primary differences between HashiCorp Vault and OpenBao:

1. License: BSL vs. MPLv2.

2. Development model: cathedral vs. bazaar.

3. Maturity: production vs. barely started.

It sounds to me like you don't care about the license, trust cathedral more than bazaar, and value the maturity of the incumbent Vault project. If you're currently a Vault user, I would stay the course for now.

I don't think there's any particular mistrust of HashiCorp in the sense that they will compromise your security, either deliberately or incompetently. However, there is an awareness that their interests aren't necessarily aligned with their customers'.

I mainly use it for text expansion, which I find easier to remember and type than multi-key shortcuts. That part of my brain is already full of Emacs key bindings. I switched from TextExpander to Keyboard Maestro years ago.

I have web-specific macros that basically act as cross-browser bookmarks. I have terminal-specific macros for the shell and other REPLs. I even have a macro for a VPN client that refuses to remember my username (over twenty characters, including a domain name).

The metadata says this was digitized by the Internet Archive in 2016. There is a better scan of a better copy of the book at the Internet Archive's library:

https://archive.org/embed/razoredgebookofs0000jura

I don't know whether the book is still in print, but it's in stock at Amazon:

https://www.amazon.com/Razor-Edge-Book-Sharpening/dp/0966605...

I use a mostly vanilla Emacs on macOS from emacsforosx.com, and I am definitely a laggard when it comes to adopting some of these modern features. I'd love to use eglot to make Emacs as smart as VS Code, but it takes some doing, so I'm still leaning on a TAGS file for navigation.

Meanwhile, each new version seems to break something in my crusty old .emacs file. My latest frustration is that drag and drop stopped working after upgrading from 28 to 29.

I don't know if a starter kit is necessarily the answer, but there's something to be said for a batteries included approach.

Yes, although Wirecutter is frequently one of the inputs into my buying decision, I find the one-size-fits-all approach annoying at times--this is the best power strip, because we believe everyone needs at least eight outlets, and twelve is better, but more than twelve is too many. Well maybe I think anything with more than six outlets is too big, or maybe a right-angle plug is my most important criterion.

I can sometimes pick through "The competition" and find something that is slightly less awesome or slightly pricier, but that section reads like you'd be dumb to buy this model instead of our top pick, which also comes in purple.

I don't think that's a safe presumption at all. Getty, e.g., claims to have found their watermark in the output of Stable Diffusion.

More broadly, it's not a settled question of what license you need. The article presents a perspective that LLMs "would turn a Food & Wine review into a simple text recommendation of a bottle of Malbec, without attribution." I doubt that anyone has a license to republish the entire archive of the New York Times without attribution.

At the other extreme, I can "train" myself on newspapers I find discarded on the subway with no license at all. I don't need to cite my sources when I state an opinion based on everything I've ever read, seen, or heard. Is that a fair analogy to apply to LLMs?

For several years, I got cluster headaches for two months every two years or so. Thankfully, I haven't had a cycle for over four years, so whatever I tell you may be out of date, and of course everyone is different. I recommend you find a good neurologist, preferably one not too old--mine kept retiring at the most inconvenient times.

Ibuprofen, Tylenol, and indomethacin have never helped me for cluster headaches, often leading instead to rebound headaches of a different kind. The three things that help me are oxygen, sumatriptan injection, and verapamil.

The number one thing that helps stop my episode is oxygen: 15Lpm or so of pure oxygen through a non-rebreather mask. A concentrator probably won't do the trick. If you have daily episodes, you can go through an MM tank in a month.

My second choice is an Imitrex (sumatriptan) injection with a STATdose pen. Pills are worthless. The 4mg dose is just as effective as 6mg, so I can take three doses if I'm having a bad day. (The daily limit is 12mg.) They sometimes make me feel funny, and the headache occasionally comes back in an hour or two.

Verapamil helps reduce the frequency and intensity of my headaches. It has a tendency to lower my resting heart rate: I was below 40bpm one morning, and I'm not a runner.

Good luck. Clusters aren't fun.

Thanks, I did not realize that. I have one hackintosh still on 10.13.

I've actually been waiting for this release to switch to ESR. The regular release has been getting updates three or more times a month, each of which prompts daily nags to update, which gets tiresome across three or four computers with a lot of tabs open. I'm hoping for more like one per month on ESR.

Last I checked, there aren't any Xeons with E-cores. However, you're right: the most direct competitor to Bergamo seems to be Sierra Forest, a chip that will have nothing but E-cores (and no AVX-512).

Although this anecdote is presented as something that actually happened, it feels like a straw-man premise. The VP "practically growled at me" and "found Anand's questions 'super annoying.'" "His assessment that Anand was a 'bad fit' was really code for 'I don't want to feel uncomfortable.'" This seems uncharitable, but if it really does describe the VP accurately as a cartoonish pointy haired boss, this isn't the article that's going to change his mind, even if the fairy tale did have a happy ending.

I generally welcome questions in an interview, but not all questions are equal. Some may signal a lack of preparation, don't seem particularly relevant, or come across as canned, like something they read in an HBR article (just kidding).

I'm no expert interviewer on either side of the table. If there's one nugget I'd take from this article, it's a challenge to "uncover capabilities, not just experience." How do I convey my own capabilities as a job seeker, or elicit them as a hiring manager?

Incidentally, the "77% of all jobs ... require little to no creativity" link takes you to a marijuana podcast, as the Martin Prosperity Institute is apparently defunct. I believe the statistic refers to a footnote in this paper (co-authored by the author of the submitted article):

https://tspace.library.utoronto.ca/handle/1807/80108

There was a distinct change to the font rendering sometime around macOS 10.14 (Mojave). I use the same 1440p monitor for computers running 10.13 (High Sierra), 11 (Big Sur), and Windows 10. I far and away prefer High Sierra to the others, with Windows 10 slightly better than Big Sur.

I don't have a citation handy, but I believe this was a conscious design choice to favor HiDPI displays.

This is perhaps the most interesting aspect of the project--using a code generator to escape the gravitational pull of CUDA. I wonder how well it would generalize to other targets.

I bought a Pico H and a Pimoroni Display Pack with the idea of making a custom clock for a gift. I didn't realize that some of Pimoroni's displays already have an RP2040 chip onboard, or in this case an entire Pico W.

I don't really need wireless, but the real-time clock would be handy.

Interesting to see that passkeys make use of CBOR. When I looked at binary JSONs several years ago, Smile was one of my favorites, but CBOR was decent too.

Damn it! I forget why I started using Keep reminders in the first place, but that's where I put all my bill reminders, including annual and semi-annual premiums.

It is beyond a cliche at this point that Google unceremoniously dumps products and features (RIP Reader and Bookmarks), but when you're creating "a single experience for managing to-dos across Google," why, oh why, wouldn't you migrate reminders from Keep? Did they forget about it until too late?

> My heart truly bleeds for companies who face the insurmountable task of printing "may contain sesame"

The reason the regulation, and the commercial response to it, is controversial is that companies cannot simply print "may contain sesame" and be done with it.

"Statements such as 'may contain [allergen]' ... can be used to address unavoidable 'cross-contact,' only if manufacturers ... have taken every precaution to avoid cross-contact"

https://www.fda.gov/food/food-labeling-nutrition/food-allerg...

This is a counter-intuitive, and presumably unintended consequence of the regulation that sucks if you're allergic, but fining or jailing executives for complying with it is silly. Hopefully, enough other companies will see a competitive advantage in retooling their processes to deliver sesame-free products.

The subheadline and four of the first five paragraphs address this:

"In 2011, Spielberg had already explained that the guns would be returning for the 30th anniversary release, explaining that he was 'disappointed' in himself."

> * Built-in integration with external KMS service

The first thing I notice is that Tink doesn't support Azure Key Vault.

I don't see any indication that Codon makes any claim to generated code. It's just that the license prohibits production/commercial use of the compiler--not much different than other proprietary compilers like, say, Visual Studio.

Another big difference: "Codon is licensed under the Business Source License (BSL), which means its source code is publicly available and it's free for non-production use. ... each version of Codon converts to an actual open source license (specifically, Apache) after 3 years."

https://docs.exaloop.io/codon/general/faq