fernly

I've done, and still do, regular blogging. I have a domain, and everything I post is mine. However it can't be "indieweb" because I used Wordpress. Hence my current blog is dependent on a corporation, as were a couple of older ones that were built in Blogger (and still exist, last updated nearly a decade ago).

That dependency is because isn't feasible for me to own my own server. If I did, it would either be virtual, in amazon's or microsoft's cloud, or it would be real hardware but still in some corporate server farm. It just wouldn't be practical for me to own (and maintain) a physical server with an UPS and permanent uplink to the net.

Says, not inflation-adjusted. With reason; adjusting those 1960-1980 prices for inflation would make the graph a lot taller.

Pricing "per GB" before 1990 is unrealistic, though; nobody thought in GB or purchased GB quantities, or conceived of GB systems. I remember a moment circa 1973 when I saw an IBM CE about to do an upgrade on a 370 system at Cal Berkeley. He had a box with several carefully-packed, large circuit boards. "So, is that a megabyte?" I asked. "Yup, that's a meg."

Per Wikipedia, it "is a colorless, odorless, non-flammable, and non-toxic gas."

When used as a contrast agent for ultrasound, it "has been used to examine the vascularity of tumours" -- which would be similar to its use in the OP. Then "[i]t remains visible in the blood for 3 to 8 minutes, and is exhaled by the lungs."

So -- not collected and excreted by the liver, as I at first thought.

Ken Shirrif recently published a survey of the "system/4 pi" family of which this is a member:

https://www.righto.com/2026/03/ibm-4-pi-computer-history.htm...

From the Conclusion section of the PDF[1],

"Multimodal AI systems are increasingly deployed on the assumption that their benchmark performance reflects genuine visual understanding. Our results fundamentally challenge these assumptions. Across every model-benchmark pair tested, the accuracy that frontier models achieved without any access to images exceeded the additional accuracy they gained when images were provided. Moreover, a text-only 3-billion-parameter model, trained solely on question-answer pairs stripped of images, outperformed all frontier multimodal systems and human radiologists on a held-out chest radiology benchmark. Taken together, these results demonstrate that high benchmark accuracy does not reliably indicate visual understanding."

Basically, they are so good at extracting clues from the text of the questions, and extrapolating from them, that they proceed to answer _as if_ they had an image to view. With confidence, of course.

[1] https://arxiv.org/pdf/2603.21687

HTML/CSS components were not available to Guido in 1988. It was in 1989 that Tim Berners-Lee defined the first version of HTML[0] and CSS was proposed five years later[1]. In 1988 Guido would have known about the Mac (1984) which he cites, and the X Window system, which was 5 years old.[2]

[0] https://en.wikipedia.org/wiki/HTML#Development

[1] https://en.wikipedia.org/wiki/CSS#History

[2] https://en.wikipedia.org/wiki/X_Window_System#History

Not that hard if you are in young to middle years and have any job experience. I asked Perplexity "If an American citizen, a trained engineer with some experience, desired to work abroad in the EU or an English-first nation, what are some good websites to check?"

I suggest you do the same -- the reply lists a dozen promising sites.

https://www.perplexity.ai/search/if-an-american-citizen-a-tr...

Set aside the effect within Wikipedia and consider the larger picture, millions of people generating text with LLMs and at least some of that text being accepted as correct by millions of readers.

The WikiEdu article clearly demonstrates what everyone should have known already: an LLM has no commitment to the truth. An LLM's only commitment is to correct syntax.

wotsit do?

Actually more fun is the "henge near me" page[1] which lets you test alignment on a map of a chosen city. Very nice interactive use of Google Maps, flipping red alignment bars on and off as the date changes.

On March 12, 2026, all the avenues in the Sunset and Richmond districts of San Francisco are Henge candidates. These are the avenues that are named in alphabetic sequence, North to South. Anyone who grew up in The City can recite the whole sequence (Anza, Balboa, Cabrillo... Vicente, Wawona, Yorba).

[1] https://hengefinder.rcdis.co/henge_near_me

agree, I feel dumb but don't see subtle issue.

Also when copy/pasting into Python to try it, I got an error because \“ is in fact U+201C not an ASCII quote. (Surely that's not the subtle issue?)

Questions while watching the video.

Calendar is central, but I use a Google calendar which is important to me. Connect it?

Seems like a "dlog" is a calendar entry. So is my "journal" broken up into separate pages, not a sequential document or blog?

2:30 ff, strongly suggest that for your next video you pre-script it to avoid fumbling and mumbling.

5:10 side note, interesting that your personality(?) model was from 2018, well before LLMs.

7:50 for an app to produce such output (impact of a friend on mood) you surely must do a copious amount of extremely frank journaling. When, and in what format? As scattered calendar entries? I'm confused how I fuel the app.

10:40 relating diary entries (reported activities and attitudes) to one's stated goals -- this is what I would expect an AI to do, and tell me about them rather than the reverse.

I'm sorry, I just don't see how I could use or adapt to something like this when I have a well-established diary/blog and calendar, it would mean changing many daily habits and adding what looks like a lot of detail work.

None of the sliders sound anything at all like my tinnitus, which is a very high complex hiss, maybe up around 6-9Khz? and steady, or varying slowly in volume. But no beeps or boops like this system.

Another nice quote,

The next logical step is to invent a way to scale linearly with the number of constraints. “That is the North Star for all this research,” she said. But it would require a completely new strategy. “We are not at risk of achieving this anytime soon.”

Some quite random claims here, can anyone provide citations?

GPT-5 can do things no other A.I. can do. It can hack into a web server. It can design novel forms of life. It can even build its own A.I. (albeit a much simpler one) from scratch

Especially "design novel forms of life" blinks "Citation required" in neon colors.

The article goes on from there with a lot of very credible, real-world jail-break examples. But that opener...

three Scottish brothers have set a new world record by completing the first and fastest unsupported row across the Pacific Ocean... The previous record of 159 days had been set in 2014 by Russian rower Fedor Konyukhov.

Um, if they are the first to do it, how can there be a previous record? I guess the Konukhov trip was "supported"? Or not "full"?

Refer to David Brooks' opinion piece in today's NYT, about Republican Nihilism. He claims there is a spirit of "burn it all down". Seems to be seconded by several comments in this thread.

Based only on the headings and titles of the charts on the first few pages? I smell an AI writing. I mean, do these titles sound like an intelligent human wrote them?

"Charts paint thousands of words..."

"Leading USA-Based LLM User" -- what does that even mean? With a value of "800MM" where MM is what units?

"AI Usage + Cost + Loss Growth = Unprecedented" -- how can you add three things that don't appear to be commensurable? Also, what is "Loss Growth" and how does it add to "Usage"?

There are plenty more examples in the charts later. The Overview section, while not so obviously AI-written, has an over-the-top enthusiasm and loose structure that makes me squeamish. I don't trust it, but that's just me I guess.

The address given for the Byte Shop, "1063 El Camino Real in Mountain View", is ambiguous. It needs to specify either 1063 EAST El Camino Real or 1063 WEST El Camino Real, two quite different locations.

Neither of those matches the store that I remember patronizing circa 1978 or so, to buy a California Computer Systems S-100 box. That would have been on El Camino just north of Grant Road, circa 80 W El Camino.

Well I just had to try it on Claude 4.0, I mean somebody has to, right? and it did a clean, if rather terse, breakdown, concluding with:

Caesar's last breath: ~0.5 liters (typical final exhale)

Total atmospheric volume: Earth's atmosphere has a mass of about 5×10^18 kg. Using the ideal gas law with average molecular weight of air (~29 g/mol), this gives roughly 4×10^44 molecules total.

Molecules in Caesar's breath: 0.5 liters at standard conditions contains about 1.3×10^22 molecules.

Your inhale: ~0.5 liters also contains about 1.3×10^22 molecules.

The fraction: Caesar's molecules represent (1.3×10^22)/(4×10^44) = 3.25×10^-23 of all atmospheric molecules.

Final answer: (1.3×10^22) × (3.25×10^-23) ≈ 0.4 molecules

So statistically, you inhale less than one molecule from Caesar's last breath with each inhalation, but over the course of a day's breathing, you'd likely inhale several molecules that were once in his lungs as he died.

Compare to Claude Projects?

https://www.anthropic.com/news/projects

Running spellcheck is a standard step on every page of proofreading. There's a "wordcheck" button in proofing UI.

Perplexity had no problem with the question, replying in part, "Paul Newman is known to have struggled with alcohol throughout his life, as revealed in various accounts, including his own memoir and documentaries about his life...."

A bit of context regarding Project Gutenberg. Its intake process is far from casual. Take a look at Project Gutenberg Distributed Proofreaders (PGDP, [0],[1]), one of the oldest "crowd-sourcing" projects on the net (est. 2000). As you can see from [0], every book goes through three rounds of proofing, where volunteers read each page of text and compare it to the scanned image; then through two rounds of format review, where other volunteers insert or review format markup.

From that 5-pass process the marked-up text is handed to a volunteer "post-processor" who assembles the final HTML or e-book file; then the completed book gets one more "smooth reading" pass before it is posted to PG.

This it the process that produces the books input to Standard Ebooks. That they can still find scanner errors ("tne" for "the", a typical "scanno") demonstrates how difficult it is to see those. But their presence isn't from carelessness or disregard for the value of the books.

In the 20-teens I put in hundreds of volunteer hours at PGDP in all the above roles, and it was very satisfying work. I'd recommend it to anyone wanting an online hobby that feels constructive. Volunteering time to Standard Ebooks would probably feel good as well.

[0] https://www.pgdp.net/c/activity_hub.php

[1] https://en.wikipedia.org/wiki/Distributed_Proofreaders

Does anyone use Google for search any more? My browser default for five years has been duckduckgo. More recently, for any factual question (as opposed to just finding some organization's website given its name, which DDG does fine) I use perplexity. Or on my phone, Claude.

I find the OP very difficult to comprehend, to the point that I question whether it has content at all. One difficulty is in understanding their use of the word "embedding", defined (so to speak) as "internal representations (embeddings)", and their free use of the word to relate, and even equate, LLM internal structure to brain internal structure. They are simply assuming that there is a brain "embedding" that can be directly compared to the matrix of numerical weights that comprise an LLM's training. That seems a highly dubious assumption, to the point of being hand-waving.

They mention a profound difference in the opening paragraph, "Large language models do not depend on symbolic parts of speech or syntactic rules". Human language models very obviously and evidently do. On that basis alone, it can't be valid to just assume that a human "embedding" is equivalent to an LLM "embedding", for input or output.

Computer History Museum blog post with pictures and background info:

https://computerhistory.org/blog/chm-releases-alexnet-source...

I would think that if you couldn't set up a modern computer so that it launched a BASIC interpreter immediately on completing boot-up, you could at least have the BASIC ready as an obvious icon on the desktop and taskbar/dock, so that a double-click would get it going. Microsoft 365 BASIC is a thing for Win11, and on the Apple App Store there is BestBASIC with an example of it drawing the Mandelbrot set. There are also LOGO interpreters for Win and Mac.

That's what I think the tech at this stage cannot do. You make two clips from the same prompt with a minor change, e.g.

a thief threatens a man with a gun, demanding his money, then fires the gun (etc add details)

the thief runs away, while his victim slowly collapses on the sidewalk (etc same details)

Would you get the same characters, wearing the identical clothing, the same lighting and identical background details? You need all these elements to be the same, that's what filmmakers call "continuity". I doubt that Veo or any of the generators would actually produce continuity.

Superficially impressive but what is the actual use case of the present state of the art? It makes 10-second demos, fine. But can a producer get a second shot of the same scene and the same characters, with visual continuity? Or a third, etc? In other words, can it be used to create a coherent movie --even a 60-second commercial -- with multiple shots having continuity of faces, backgrounds, and lighting?

This quote suggests not: "maintaining complete consistency throughout complex scenes or those with complex motion, remains a challenge."