atombender

There are plenty of signs if you look for them.

- "But the same closure that empties a granary fills somebody’s brokerage account." Claude is obsessed with comparisons linked by "the same". There are several instances, like "Poland made the same swap, with oil" (but that one is odd because "the same" isn't referring to anything).

- "Europe’s potato belt tells the story in a single season": Claude also loves this "noun verbs the noun in a single noun" phrasing.

- "A country that loses its regular supplier does not only pay more. It ends up dealing with people it does not know." Claude's classic "it's not X, it's Y" constructions.

- "Strategists do not describe a stampede home. They describe Japan quietly ceasing to be the buyer everyone could count on, and long-term borrowing costs rising everywhere as a result." A variant following the pattern "it's not X, it's Y, and Z as a result." This is an interesting pattern, one that I don't think I've ever seen a human write. A human would likely write "...leading to rising long-term borrowing costs" or similar.

- The use of a verb (rising) there in the third part is typical Claude, as it really likes to verb when a noun would read better. Other examples: "markets price another rise"; "volume freezes, prices stay put and sellers wait"; "after five years the rate floats".

- "The test Tusk describes is a test of belief, and it arrives just as the Pentagon decides how many Americans stay." The juxtaposition here (a test that arrives) is very odd and typical of Claude.

- What's also generally striking is the proliferation of "X becomes Y" phrasings, which is something Claude loves to do. Price shock turned into empty pumps. A story is told in a single season.

I'll stop there.

What's fascinating to me is that I personally find this kind of text exhausting to read, but apparently other people don't notice it at all.

If we ignore the AI stuff, I will still say this is not good writing. What's so grating and noticeable for me is the monotonous style. The short, clipped sentences like little stabs of information. It never varies; the tempo is just fact, fact, fact. It's absolutely exhausting to read pages and pages of this.

Good non-fiction writing is enjoyable to read for a myriad reasons; authors use metaphor, juxtaposition, tempo, humor, irony, allusion, and all sorts of techniques to make an article enjoyable to read. Just throwing down facts one after another with this kind of breathless, staccato noir voice is not that, in my personal opinion.

I don't know how much AI is to blame, maybe the Polish version is different or if this is a deliberate style.

Not often, but just because it's rare doesn't mean it's not part of my vocabulary! I know a lot more words than I actually use in daily communication, and I suspect that's true of most people.

For example, I have never used (as far as I remember) "inchoate", "protean", or "exegesis", but I know perfectly well what they mean.

> hold your horses to paint it as dirt cheap

For a moment I thought this was going to be a metaphor — maybe an ancient Chinese proverb about how paint brushes are made from horsehair and how you can't hold the horse to paint before you've turned the hair into a brush.

I guess it depends on your background. For me, those words are part of my vocabulary, and I would not think twice when seeing them in an article. The author might not have been thesaurus-hunting if they had already internalized those words, right?

However, I agree that some authors go too hard on the thesaurus for seemingly no good reason. The New Yorker and certain other "high brow" publications are well known for this. "Obstreperous", "prelapsarian", and "fuliginous" are not, I would argue, part of anyone's vocabulary.

Does Zen have multi-account container tabs? Edit: Yes, it does! [1]

(I need them partly because of Google, as Google has a broken login system. I use GGP at work plus a bunch of work apps that use Google for SSO. But for some products like Maps and YouTube, I use my personal account. Doing that on e.g. Chrome is painful, because every URL opens with whatever is your default login, and whatever I set as the default, it's usually wrong.)

[1] https://docs.zen-browser.app/user-manual/workspaces

Does nobody read the articles? The article is about a Victorian reproduction in Reading, England.

Claude-isms like this is not what humans write and it makes my spidey sense tingle unnervingly:

> Plain Markdown on your own disk, nothing to sign in to, and the whole dialect when you want it — five things worth knowing first, each argued in detail further down

Just to break it down:

"Plain Markdown on your own disk": What an odd way of saying it stores Markdown files on your computer. In this day and age, "Stores all your data locally" is an easy way say this.

"...nothing to sign in to": Reads so awkwardly, I think, because it declares that something that doesn't exist is a good thing. It's like advertising home delivery with "no store to go to". I would put "no account needed" or "no sign-up required".

"...and the whole dialect when you want it": The "when you want it" is so weird here. Why not say it supports all of Markdown?

"...five things worth knowing first, each argued in detail further down": There is something uniquely strange about this phrasing. Why not just say: "Here's five things you should know"? Or just drop it. It doesn't add anything.

Claude is terrible at writing copy. GPT is a bit better. At the moment, Gemini is the best one, I find.

Would also love to read a write-up about this!

The readme says:

> Mermaid, Graphviz and D2 let you declare boxes and connections, then determine positions for you. If you have a particular picture in mind, these aren't the right tool.

D2's layout engine TALA [1] actually allows you to provide a lot of manual control. TALA used to be commercial, but is now open source.

In my own testing, TALA tends to produce much better layouts than Mermaid, including the ELK layout algorithm (which is not Mermaid's default; D2 also supports ELK).

TALA supports controlling direction per container, defining which shapes have to be near each other, and so on. I've not played with this yet, though.

[1] https://d2lang.com/tour/tala/

My understanding (possibly flawed) is that Stonebraker isn't directly involved anymore.

The original idea that he worked on with the DBOS people at MIT and Stanford was very different and much, much more ambitious, which is why the name DBOS seems a little out of place now. The original idea was much closer to a "database OS".

Here [1] is the paper, which proposes that "To improve the scalability, security and operability of OSes, we propose a data-centric architecture: designing the OS to explicitly separate data from computation, and centralize all state in the OS into a uniform data model. In particular, we propose using database tables, a simple data model that has been used and optimized for decades, to represent OS state. With the data-centric approach, the process table, scheduler state, flow tables, permissions tables, etc all become database tables in the OS kernel, allowing the system to offer a uniform interface for querying this state."

The team later published another paper based on their prototype work [2].

Instead, they basically implemented Temporal as a client library with Postgres as the state layer. It's good, but only tangentially related to the original vision.

Maybe the long-term plan is an actual database OS, but it kind of looks like they decided they had to pivot to something much simpler, and slapped on an "for AI" like everyone is doing these days.

[1] https://arxiv.org/abs/2007.11112

[2] https://dl.acm.org/doi/10.14778/3485450.3485454

Astra is better here, but the one I'm the most impressed with is Gemini. It's always been good, but 3.6 Flash is even better. It writes in a pleasant, human style. Not perfect, but it has a good balance between technical accuracy and readability that is better than what I've seen from any other mainstream model.

Their comments are also AI-generated, for some reason.

I use Jemalloc on a project because it has per-thread allocation counters. This lets me not only track what each thread is using, but also restrict workloads by enforcing a memory budget. Since my application is almost completely CPU-bound, I allocate one thread per core and then have some smart scheduling to route requests to the threads. I looked at tcmalloc and mimalloc, and neither had this feature at the time, nor did they seem to have any similar feature that could let me have per-thread heaps, which surprised me.

"brew install postgresql", unless you meant without Homebrew, in which case you could download the sources manually and compile with "make".

Replace "need" with "want" and it might make more sense?

Homebrew is the Mac equivalent of APT, Pacman, RPM, BSD ports, etc. It's a fairly traditional package manager for software that doesn't come with macOS. "brew install postgresql" installs Postgres into /opt/homebrew, for example.

If true AGI superintelligence is invented, I don't see how it will be made available to the masses, at least not intentionally. There will be no incentive for the likes of Anthropic and OpenAI to give such immense power to anyone else. Whoever has the superintelligence will be able to race ahead of everyone else.

However, such technology may be leaked, or it may (as what happened with LLMs) be so simple to replicate that anyone can do it. That doesn't mean it will; for example, LLMs are possible thanks to the affordability of GPUs, but they're only affordable right now (and increasingly less so) because of free market economics.

We are in the honeymoon phase of some tech that's in its infancy. The "democratizing" aspect of AI won't realistically last, I think. The current non-AGI AI won't necessarily go away, but it will be rendered obsolete.

No, the browser version of Plex is much simpler because you can just send a link. The receiver only needs to go through a simple "type your email and password" sign-up process, and that's it. Nothing to install, works everywhere.

Tailscale is exponentially worse, though. Not only would I have ask them to go to Jellyfin and sign up and log in, I would have to explain that they can't do that until they've installed the Tailscale app (which I have to explain because most non-techies do not know what a VPN is). Tailscale requires its own account, so they have to first sign up to that, and I have to add them as an external user first.

Like you, I have weird ear canals, and I've always struggled with fit; the Apple ones are the worst.

I have the Sony earbuds, and while I agree they fit a little better, I find the memory foam disgusting. It also degrades over time and becomes brittle. My solution was the Alza SednaEarfit Max, which are silicone earbud tips that fit the Sony. They are easy to clean and come with the tiny mesh that prevents ear wax from getting into the electronic part of the earbud. Of all the tips I've tried, these stay the best.

However, they still fall out too easily if I'm eating (from jaw movements) or lying on my side.

PS. You mean Sony WH1000MX4. Not Sennheiser. Although Sennheiser makes the Momentum True Wireless, which comes with silicone earbud tips.

Is there a way to test a cable to see if it's PVC and not silicone?

I don't believe in Jellyfin being secure enough to put on the public Internet. In my case, Jellyfin is on my home network, and the box has access to a lot of files and devices.

I think you overestimate the technical capabilities of my family or their eagerness to install software. When I ran Plex, and nobody ever installed the Plex app like I recommended to them, even though it's one click away in all the app stores.

I don't trust Jellyfin to be secure enough to be put on the public Internet.

I left Plex for Jellyfin + Infuse a couple of years ago, and I'm not looking back.

But the sore point right now is outside access, which Jellyfin really isn't made for. I set up Tailscale so I can connect Infuse to Jellyfin remotely. But that's not exactly as smooth as Plex. I can't give friends and family access, either, which was dead easy with Plex.

In terms of the UX, Infuse isn't very good, but it does the job okay. The built-in subtitle fetcher is amazing; that stuff never worked for me with Plex.

Gibson is one of those writers who are much better prose stylists than plotters.

I enjoy reading his prose immensely, but he's been stamping out books with the same template for ever: an expert is hired to do something. Turns out there's a man with unlimited cash flow funding everything, but was hiding his true intentions (usually in order to lay a trap), which are revealed, and then some McGuffin is saved or eliminated or whatever; meanwhile, the main character mostly didn't do anything, just a passive spectator to some plot unfolding around them. The one I will forgive most easily here is Count Zero, which I think is much better than its reputation (except for the Haitian voodoo deity stuff, which didn't land for me); while the template is the same, the ruse connects well with Neuromancer, and it's just a fun, action-filled book that is definitely not boring. (The ruse in The Peripheral, by comparison, was deeply unsatisfying. Agency was junk, most of the book describing preparations for a military attack while the main character sleeps. Even the prose was much worse and much sloppier than usual.)

What's interesting is that his early short stories are much more original, I think. Burning Chrome, New Rose Hotel, and Johnny Mnemonic, for example, show flashes of originality that maybe Gibson wasn't able to keep up when plotting out whole novels.

The standards of purity have changed over time, and most of the US gold dates from before the 1930s; and so a lot of US-domiciled gold have different purities that don't follow the current international standard. The purity requirement has been going up over time. A lot of US gold is 90% gold and 10% copper, but gold traded on the international market needs to be at least 99.5% pure. Here [1] is a good breakdown.

[1] https://www.bficapital.com/blog/what-is-inside-fort-knox

About the Langlands program, Nunberphile has an excellent episode with Edward Frenkel explaining what it's about: https://youtu.be/4dyytPboqvE.

The staggering number of fonts used on that award [1] is something to behold. I count maybe six?

[1] https://images.squarespace-cdn.com/content/v1/669edd08b2a3fe...

Does anything like this exist for macOS?

Head's up that this adds a non-trivial amount of CPU load when the screensaver isn't running. On macOS 26, the "legacy screensaver" process takes about 30% once this is enabled [1].

[1] https://i.imgur.com/vcrq7mG.png

Norton Speed Disk [1] for DOS was more satisfying to watch because it would visualize reads and writes. So you'd see it grab blocks and then smush them together into contiguous areas. The Windows version was not, as I recall, as satisfying.

[1] https://youtu.be/-T712Wbc6a0