It's 1965. You're watching your favorite television talk show. They've just started broadcasting in color, and you've got a color television set, so you're pretty excited. At the appointed hour, the jazzy theme song plays, and the host comes out and begins to deliver his monologue. A few topical jokes at first, which aren't your favorite, but they're amusing enough: a joke about LBJ's twangy accent, a bit about women in miniskirts driving on the Interstate Highway system while listening to the Beatles, a pitch-perfect imitation of United Auto Workers president Walter Reuther ordering dinner at a fancy French restaurant. Following some banter with the bandleader, the host sits down at his desk and picks up some envelopes. "And now," he announces with an impish grin, "we come to the part of the show where I read your letters."
You haven't sent any letters to this talk show host. Are you confused?
I didn't see the two comments that replied to yours but I'm interested in what they said. The OP deletion that didn't reply to you is pretty easy to guess but it sounds like the two replies to you were maybe pretty spicy so I'd love to hear them if you remember, or at least a paraphase if they can't be printed here
This is a cool history story and I liked the writing. But the review sort of lost me in Part Eight, where it showed me a screenshot of a Xanadu browser, and it asked: "Why shouldn’t the internet look (and work) a little more like this?"
It's possible that this is the sort of thing that you have to use for ten minutes to understand. But it looked to me like a normal (image-less) browser with a bunch of tiny unreadable slices of text on either side, and I don't feel any particular need for the internet to look more like that.
I would agree with the author that there are a lot of problems with the internet, but I'm left with no idea how Xanadu thinks it's going to solve them.
Concretely: Xanadu offers "connection, accountability, verifiability", but it doesn't explain how it's going to get any of those things.
Verifiability: Is this going to be one of those Internet things where you have to upload a photo of your drivers license before you can post? If so, does it offer protection against getting doxxed? If not, how is anything verifiable?
Accountability: If you post something deliberately wrong on Xanadu, what specifically happens?
Connection: Can a spammer create a page about their new memecoin, and "connect" it to the most-used page on the Internet so that everyone in the world sees it? If not, why not?
There are a lot of people that want connection, accountability, and verifiability. But it's not enough to handwave saying "this project solves those problems". These are actually-hard problems and you have to explain how you solve them.
There is an extension "ACX Tweaks" on the Chrome Extension store that modifies the styling on this site in a number of user-configurable ways, including "Show hearts". There may be similar ones for other browsers. Enough people are using this that I can tell the two parent comments have 17 and 15 hearts respectively, and yours has three.
This. The whole project strikes me as something which has to be proprietary and controlled by someone. The web we have is decentralised, because it basically lets anyone who can rent server space put whatever they want in a window of whomever’s willing to open it. Xanadu seems much more restrictive; you could build Xanadu and stick it on the web, but I don’t think you could build the web and stick it on Xanadu.
(Technically, there is also some centralization on the web, for example until a decade ago the 13 root DNS servers were all controlled by ICANN.)
Yes, Xanadu would either need a centralized database (like Wikipedia) or a way to make multiple databases cooperate (like Fediverse), which opens another bag of problems.
If I am following the description (no guarantees there) Xanadu could be the web. I think it works like this: every time someone wishes to borrow an idea from one piece of media, a link is created back on that original piece of media with a link to everyone who has borrowed anything from it. Simultaneously, there is a link back from the new article to all the originals it borrows ideas from (I don't know how you stop people from claiming ideas they didn't originate, but the current web can't do that either). In theory, one could read an article, track back to where it got an idea, see everyone else who borrowed an idea from that, and go merrily on. It's network of ideas, not just documents.
I don’t know pubmed in particular, but many academic paper databases now have a page for each paper together with links to click to see all papers cited in this one, and all papers citing this one. The latter isn’t always complete of course - at best it’s all papers published in journals and volumes indexed by this site. But that’s often pretty good.
They show somewhat different numbers of citations (175 vs 110) but I can’t tell how many are missed citations at one or duplicates at the other.
This is the sort of mechanism I advocate to grad students who want to learn some literature, and ensure that their response to a paper is somewhat original.
What’s to stop them not putting the link in though? Or is this just the web but with very strong norms of putting links in (and where links are somehow reciprocal)?
I assume the process is automated somehow. That is, if you copy something off someone's document, the system knows this and creates the link back to yours.
So a sufficiently popular piece of original media will be seen as 5% original content, 95% links to everyone who ever riffed on that content? And if the original is sufficiently controversial, I suspect half of those links will have labels that are cheap little attacks against the original with the promise of more to come if you click on the link. Outrage galore, and DDOS to useful communication.
Yeah, it's not clear to me from the description where the tracks are located, but I assume it's all on the servers somewhere. This is supposed to be a distributed computation model, so it could be anywhere, I suppose. Were we doing this now, some sort of voting system might be put in place, like in Wikipedia.
The cypherpunks of that era had some thoughts on deterring spammers from bringing down their anonymous email/remail systems. Hal Finney's Reusable Proof of Work was intended for that. Now we have Bitcoin and other crypto systems using distributed blockchains, but spam is handled by AIs looking for suspicious indicators rather than requiring that any user pay or verify somehow that they aren't a bot.
Yep, completely agree. This Xanadu thing seems like a really lame how-to-do-more-of-the-same-just-in-a-different-medium solution. It reduces the time and effort for a scientist who reads a paper, notes an interesting reference in another journal, and then goes to read that journal. Or the newshound who reads a piece of news, spots an interesting reference to a piece of history she doesn’t know, and looks it up in an encyclopedia. It solves those specific problems, but it doesn’t do two things that it claims to do:
1) It doesn’t really reflect the “variety of concepts and concept structures in a way that makes them maximally available and useful to the needs of the human's mental-structure development”. People make all sorts of weird connections that cannot be represented by this very dry source-of-quotation structure. When I read papers, I think of poems; when I read poems, I wonder about crystal structures. The weirdness of the existing web actually reflects the weirdness of people’s brains much better.
2) It fails to respect the fact that the web really is a different medium. It’s just an attempt to take the professionally-written journal and encyclopedia and digitise it. But the web really allows anyone to write, and accepts the fact that most of us are terrible writers. Terrible writers and videographers need very loose structures to contain their blatherings, and we got it.
> Can a spammer create a page about their new memecoin, and "connect" it to the most-used page on the Internet so that everyone in the world sees it? If not, why not?
Yeah, that was my first thought after reading the proposal of bidirectional hyperlinks. They are great if both sides agree that the two documents should be linked. It is a problem when one side wants to make a link, and the other side does not want to make a reciprocal link. Current web solves it by links being one-directional, so everyone decides what their document links to.
Even assuming good faith, some documents are just vastly more important than others. Imagine millions of documents linking the definition of atom in Wikipedia. Should the Wikipedia article on atom link all of them? (At the very least, we would need a system to organize the outgoing links, maybe group them in categories such as "textbooks referencing 'atom'" and "personal web pages referencing 'atom'" etc.)
But maybe this is too much to expect from version 1.0. The e-mail protocol worked well at the beginning, and remains kinda usable today, despite being completely unprotected against spam. Maybe if bidirectional links became the norm, we would figure out some solutions to abuse afterwards.
One possible solution would be that the bidirectional hyperlinks need to get the target's approval, and in the meanwhile act like unidirectional hyperlinks. A mechanism similar to "friend requests" on Facebook would show you "hyperlink requests" and give you the choice to approve or ignore. (Of course, there is still potential for abuse, for example the other party might change the contents of their page after you have approved of the link.)
So, maybe something like this could start in a high-trust environment, and gradually adapt to less and less trustworthy users?
>Even assuming good faith, some documents are just vastly more important than others. Imagine millions of documents linking the definition of atom in Wikipedia. Should the Wikipedia article on atom link all of them? (At the very least, we would need a system to organize the outgoing links, maybe group them in categories such as "textbooks referencing 'atom'" and "personal web pages referencing 'atom'" etc.)
Ouch! Excellent point! My immediate worry was about the analog of distributed denial of service attacks, but, even with good faith actors, just _organizing_ the backlinks is, as you said, a problem.
>One possible solution would be that the bidirectional hyperlinks need to get the target's approval, and in the meanwhile act like unidirectional hyperlinks. A mechanism similar to "friend requests" on Facebook would show you "hyperlink requests" and give you the choice to approve or ignore.
Hmm... To first order, yes that gets around the DDOS analog. To second order, though, in addition to the problem that you already articulated, there is the problem of now needing a (human?) decision for each of those millions of links to popular pages. _Maybe_ that could be automated with today's AI, but I suspect that there is a broad set of vulnerabilities...
That sounds like the problem Bush was trying to solve. You have a vast mass of documents, how do you organize them? Well, so far as I can tell, the user creates a file that organizes them in some way useful to that person. Again, so far as I can tell, none of these guys gave a thought to anyone *sharing an organizing framework* When you look at someone else's file, all you see is your connection to it. But how do you get to the next file/book/idea? There doesn't seem to be a system for that, unless I missed it.
Seconded! In addition to your concerns about whether there is any way to actually implement verifiability and accountability I'd add:
If every link creates a backlink (stored on the linked page????) don't we automatically get a version of distributed denial of service attacks that chews up massive storage on the target page?
I think that, in our actual time line, the invention and deployment of search engines does most of what was suggested by Xanadu and lacking from just the linking technology of the world wide web - yet reasonably feasible.
In general, I think that this review (and, frankly, the architecture of the world wide web itself) takes too little account of what to do about malicious actors.
Orthogonal to that, but still an unsolved problem, even in the _absence_ of malicious actors, I think the promise of links (even very sophisticated ones) has been oversold as improving intellectual discourse more than they really can. (ignoring AI for simplicity) A rich set of links plus search capabilities gives, roughly, an exponentially increasing number of documents with search depth - but human bandwidth doesn't go up. Most arguments are, of necessity, unanswered, most links untraversed, most ideas ignored.
My interpretation (IANAP) is that with a Xanadu system you see the source of every piece of text automatically. Nothing can be misquoted, and manual attribution is unnecessary. I cannot tell if this is either practical or desirable, but it's interesting.
I actually think there are some reasonable answers to all of those points given *just* the hard-linking! Having not used the software but to take a charitable view of what was presented in the review:
Verifiability: You take verifiability to mean verification of identity and (possibly?) assume that also helps some verification of claims. I think Xanadu style linking deals more directly with the latter. If I claim 'Bob wrote X; how dare he' then within a Xanadu style system presumably I would be expected to link to verify my claim or whoever reading would reduce their credence in that claim. Of course, I can make greenfield claims without linkage "X causes Y but you won't hear that from Z because they're lying to you" but clearly we are still in a better situation than where we are now where 'Z' is often misquoted and framed without easy enough verification for the average reader to check in the middle of reading.
Accountability: Don't you think that if every deliberately wrong post had a list of indexed links to every debunk article accessible on the same page of your browser you would have a harder time convincing people of deliberately misleading things? Wouldn't you lose credibility otherwise? I'd say that is a system of accountability. Again, not saying it's perfect, but seems like a clear argument could be made in favor of it being better than what we have now and that's the bar to pass. Kinda, in a way, a twitter notes for the whole web.
Connection: Yes, a spammer could reference a popular page and get linked to it just like a spammer can currently reference a popular topic to show up in search results on that topic. Presumably the linking system would employ some type of page rank so that more pertinent links float to the top of the associated link list over the crypto scams.
> Accountability: Don't you think that if every deliberately wrong post had a list of indexed links to every debunk article accessible on the same page of your browser you would have a harder time convincing people of deliberately misleading things? Wouldn't you lose credibility otherwise?
No, not at all. It's already the case that many of the posts I read online are debunked by other posts I read online. This looks like anything from the several ACX posts where Scott responds point-by-point to a debunk article by someone else about an earlier post of Scott's, to some random person retweeting something I've never seen before purely so they could tell the world that the author of the original tweet is a Nazi or a retard or a retarded Nazi. The existence of some post B purporting to debunk some other post A gives me basically zero information in and of itself about whether post A, post B, both post A and post B, or neither post A nor post B are correct/interesting/worth my time to read/etc.
Encoding responses to posts into the system that displays posts to a reader doesn't seem drastically more innovative than e.g. the tumblr reblog system. It certainly doesn't seem like it would meaningfully change the dynamic that anyone posting on the internet about anything could be an insightful genius, a dumbass, or a spammer; and there's no way to determine that without actually reading their post critically and thinking about it.
I rather suspect that illegibility to the proles is a part of the point - the fantasy of some return to pre-Eternal-September Internet, not that most of the people talking about Eternal September would even be able to have any personal experience of pre-Eternal-September times (I certainly don't!)
To me it looked pretty intuitive. The current model of writing a webpage is (roughly) that you of sit down in front of a blank screen and typing your text document, when it's finished you hit publish and it's available for all eternity. The analogy is to writing a letter on a sheet of paper.
In this editor, you have multiple documents open and you can copy from them into what you are working on. The editor tracks this (those are the tiny texts on either side) and presumably if those are edited then so to would your document. There's no issue with citations - the format itself tracks where things come from, and the relations between them are as important as the content itself.
Maybe it helps that I've been reading postmodern fiction? Flann O'Brien borrow characters from other works because he claimed there were enough fictional characters already in existence; and one of his novels contains adverts and translations of poems as insertions. Borges played the same games with authorship and citation and etc.
I'd have liked to see an analogy in the review between the classical view that information in the brain is (somehow) stored in brain cells to the modern view that it's stored in the connections between neurons. The connections between texts being at least as important as the texts themselves seems to me the underlying idea of Bush and of Nelson.
1. Guy has cool conceptual idea but no real technical knowledge about how to implement it
2. Lots of folks with actual technical knowledge try to make it happen on and off across multiple decades
3. They fail
4. Isn't it tragic that we somehow don't have this thing that would be so awesome?
To me, (4) should be "So obviously this thing had fundamental underlying technical issues that made it unworkable". But that idea wasn't really addressed in the review, which I find mildly disappointing.
(I did enjoy the read. It's an interesting story and history!)
Honestly that's even too charitable. My takeaway is "a part for giving people lots of warm fuzzy feeling about the power of liberating information and so on, there is no actual concept there. It's just a guy who thought how awesome it would be to have awesomeness, and somehow his charisma and eloquence made some otherwise intelligent people follow him in his quest to find the phoenix he could not even properly sketch"
It sounds to me sort of like a gigantic, open version of the old Apple hypercard system. Which was great for its time, and still sort of cool in an artistic way- you can throw together stacks of media, and hope for creativ inspiration to strike, like a non linear powerpoint presentation. Sort of like a software version of a hedge maze or Chinese Scholar's Garden, the nonlinearity is the point.
But then, those only work because they're closed off. I have no idea how you'd make a version open to all publishing like the WWW. But maybe it could work on a much smaller scale, with central control, sort of like a museum that's open to all visitors but only the central staff can publish there. And then the visitors can wander around at random, having their own unique experience, without the spam and junk of the general WWW.
Tim Berners-Lee went on to propose and advocate the so-called Semantic Web, using RDF and OWL, which somewhat relates to Nelson's ideas. Although RDF and OWL are sometimes used for niche applications, it's safe to say that the Semantic Web never took off, in spite of full backing from the World Wide Web Consortium (W3C).
Perhaps the real lesson here is the bitter lesson. All of these rich hypertext approaches, whether Bush's, Nelson's, or Berners-Lee's, rely on sophisticated formalisms that require a degree of human effort at write-time. (OWL, for example, was derived from knowledge representation formalisms used in symbolic GOFAI.) Maybe that's just a mistake. Maybe the way to go is with general learning algorithms that compress knowledge into models. Maybe GPT-5 is better than what a fully developed Xanadu ever would have given us.
I think the problem may be harder than that. As Bertrand Russel said somewhere, acts of thought are like cavalry charges in battle, they're expensive and you can't do many of them. A device that makes analyzing arguments, understanding their history and all of their connections ... are pretty much like good college courses, they take a lot of work, they aren't quick, there's not that much demand for them, and 99% of the time, you don't need them.
I mean, the panic-du-jour of the Internet is that too, too many people type a question into Google and take Gemini's two-sentence answer at face value. And because they don't even bother to click the link that Gemini provides (to verify that Gemini didn't screw it up), whoever is getting click-paid for the page Gemini extracted the answer from doesn't make any money off it.
Or I could phrase it that Xanadu is *great* for people who routinely work at the highest levels of cognitive complexity. But most people never do that, very few people are paid for it, and a lot of people aren't capable of it. In the end, *What is the target market for Xanadu?*
A stunted version of the Semantic Web has been implemented, and it is taking off like crazy. It's called the "Model Control Protocol", and it is basically how some software system describes itself to an LLM so that the LLM does not hallucinate nonsense when generating content for that software system. Example: the MCP for Python allows an LLM to generate code that is syntactically correct Python.
I was wondering if that was the point the piece itself was reaching towards. Particularly with the idea that human thought is fundamentally associative.
I think Gwern has some insightful commentary on why Xanadu failed: https://gwern.net/xanadu I would summarize it as:
1. Most of what it's trying to do, beyond what we got, just isn't that useful,
2. Its ideas about how to handle copyright were unworkable,
3. A public version would have had serious problems with spam and abuse.
Really, I'm not sure I buy this idea that this was all somehow contingent on Engelbart being discredited -- is it even really true that Engelbart was discredited? That's certainly not what I'd gathered from what I'd read elsewhere... it looks to me like there's so many reasons for its failure (the reasons Gwern mentions, the fact that the people making it didn't actually have a clear idea what they were making) that I'm not sure things could have gone hugely differently here.
Gwern explicitly writes: "Xanadu wasn’t the victim of 'Worse is Better'; it was just a solution in search of a problem." And makes a number of arguments to this effect. Worse-is-better isn't an iron law of nature; C++ and Java have had plenty of market success.
I was very confused by the linked page, which seems to expect the reader to have a high level of context about programming language and operating system design.
Quote: "The concept known as "worse is better" holds that in software making (and perhaps in other arenas as well) it is better to start with a minimal creation and grow it as needed."
One question I have: Gwern writes, "Like Douglas Engelbart, Nelson enjoys (if that is the word) the honor of being a living embodiment of Cunningham’s Law: becoming one of those historic figures whose importance was that they were *so* wrong on *such* an important topic they helped popularize that they inspired others to become right."
Do you have any idea what he's talking about re: Engelbart? Is there an important respect in which the Mother of All Demos was conceptually flawed? I vaguely recall hearing allegations that Engelbart had some unworkable-in-retrospect Bush-esque ideas about how it should be used, but I can't find details and the demo looks fairly similar to how mainstream computer UIs ended up actually working (at least once Google Docs made multi-user editing a thing).
Adding my thanks for linking this - very interesting. I am curious that Gwern doesn't discuss or even mention scientific research as a use case; if I've understood this review correctly, that was the original motivation for the whole thing.
(Much) more speculatively, perhaps Xanadu feels more useful to people who process visual information in an unusual way?
I disagree with these general points (although admittedly I have also dreamed up ideas similar enough that people pointed me at Xanadu):
1. I think there's lots of information that could be more useful with a bit more than just text. It's really common when researching controversial stuff, particularly in politics, to have to follow claims through lengthy chains of links only to eventually figure out some part of it was bullsh** or it was all some telephone game. I'm convinced we can do better.
2. Lots of stuff isn't copyrighted and that Internet does fine with a wide mixture of stuff with different licenses.
3. This is a problem, but also it's a problem with tons of existing systems. I think a richer system is going to be no worse than social media (looow bar, I know).
[EDIT:] Reading through Gwern's writeup I see I misunderstood the second point, Xanadu was trying too hard to SUPPORT copyright. This is a reasonable critique. I also would say I think Gwern makes the case that in practice the UI they built for Xanadu was kind of terrible, which is a real death knell if you want to get everybody to use it.
I'm writing out of ignorance here, but one question that I don't know the answer to about Xanadu was: Is there an easy way for the user to know the cost of a payment to a copyright holder _before_ they access the copyright holder's content? One thing I find irritating about substack is that it isn't clear how much a subscription will cost before subscribing (though I may have overlooked this - but not for lack of searching for it). This can make or break a payment scheme.
As an (advanced) end-user lived and worked through the computer revolution from 1977 until the World Wide Web revolutionized the Internet (ca. 1996), I don't remember that Engelbart was either discredited or marginalized. As best I can remember, he was respected in the personal computer computer community (even though he didn't like the idea of standalone personal computers) and then the Internet community. I remember lots of articles in the popular computer magazines praising him as a visionary — especially when the graphical interfaces of Lisa and then Macintosh started blowing people's minds. Englebart was credited with the vision that created the first GUIs.
Nobody knew about the EST nonsense; if they did, it was just some goofy thing a lot of people did. Other such cults were much much worse because they turned rancid and violent- Synanon, Scientology, the Rajneesh crew in Oregon.
Yeah I've only heard about him in reference to Mother of All Demos. I've only heard that EST things very very recently, maybe because this review (while still in Google docs?) triggers an osmosis that led to Reddit?
Ok, now I have a better understanding: cancel culture was not a thing at the time. We did not judge people by cooties. There was probably no newspaper or magazine article about Engelbart that mentioned EST.
While the contest review made me interested in this Project Xanadu thing, it also left me a bit confused about how exactly Xanadu (or memex) would be different from WWW. It was Gwern's essay which cleared up the confusion and also acted as a real review (i.e. saying whether Xanadu was good or bad).
I'm going to digress here and saying that this comment makes me actually see what gwern is, when before I just encounter their random articles linked here from time to time. And I can only say that, how could I not discover them much earlier than this? Even on this topic about Xanadu, I find their article much more competent and have personal touch. Especially because they actually seem to have tried to make something similar, resulting in their website having the actual implementation of transclusion that I actually like and hope is implemented everywhere else. ACX should go off Substack and hire whoever is making gwern's front-end.
The design article makes me explore their archives and discover that they write about things that are actually much much closer to my own interest, especially anime and Haskell. Rationality and psychiatry is nice to read sometimes, but nothing beats discussing Umineko, GAN, and Cabal. Looks like I found myself but better in everyway.
Even though i may not agree will all that they write and their writing is still not as excellent as SSC/ACX (well, it's a high bar), I think I may found a new blog to binge and it's all thanks to this comment.
I'd say that the biggest weakness of gwern.net is that it doesn't facilitate any kind of commenting/discussion. It's his own personal archive that he obscurely tinkers with. Sometimes it gets linked to, I enjoy reading whatever was linked, but with no way to engage it's not enticing to dive deeper.
Small but kind of glaring error: the proximity fuse wasn't an anti-tank weapon in World War II, it was mostly used against air targets and secondarily against infantry. When you're trying to destroy a tank you want your shell to be intact when it impacts so it can penetrate the armor, not explode next to the target.
This is general for many of these reviews, but I feel a lot of them aren't even reviews, more like a summary of some interesting history. There's subjectivity at what, two points ("Xanadu looks cool" and "Nelson seems annoying")?
many book reviews are reviews of histories or biographies, and are mostly summarizations of the stories within, sprinkled with some author commentary, who hopefully drags in other points of view, etc. this review seems like mostly a fairly standard book review of Nelson's autobiography, in this way, and i think if it were presented as such it wouldn't stand out as not being a "review"
True, but this seems to be what people want. There were definitely actual reviews in the submissions and it doesn't seem like they were extra-likely to be chosen.
The review contest has always been an essay contest, much like the question time after presentation or a panel tends to be an "impromptu public speech about some preferably related topic" hour.
This is, of course, even moreso when you're not talking about something like a book that someone might, you know, want to read on the basis of the review (it's not like someone's going to be able to try out Xanadu-as-fully-imagined-by-Nelson on the basis of this review, after all)
I agree - the most highly regarded reviews appeal to the interests of the readers here and mimic the style of the substack reasonably closely. It makes sense: they come here to read text in a particular style on a fairly narrow range of topics.
Particularly with finalist reviews (of books) in the last year or two, none have motivated me even slightly to buy or read the book. (It could be argued that this isn't the goal of literary criticism - neither are the essays going to have much impact on authors and academics and future books, I expect.) So as reviews they fail, but then Scott is not in the business of selling books. As essays they succeed very well. They give an overview of a general area, normally with reference to what I assume is a reasonably good book, and maybe some insight into how thinking in the area falls into line with a rationalist world-view.
I have a little bit of a personal connection to Ted, as I was his neighbor in Sausalito for a bit. I moved onto a ferry boat called the Vallejo in 1994. You can see it clearly in this home video walkthrough by Ted and his son, near the end. https://archive.org/details/SouthFortyPier1988
He may have even come over for one of our infamous parties.
Not only is Elon not insane, but he is one of the most accomplished human beings on this planet, and in fact in history. Putting him alongside Edison, ... is entirely appropriate.
he seems pretty erratic lately with a some addictions and obsessions that seem to be destroying his health and wealth. whether he's a very accomplished person has nothing to do with it, unless you think that success immunizes you against mental problems
if you don't want me to call these things addictions im happy to use whatever word you prefer but it wouldn't change the underlying fact that there are many reports that he is using ketamine frequently from people who know him personally. you can see how much he sleeps based on what times of the day he tweets: for years he was getting only a few hours of sleep per night.
Yes, I'm assuming a man that runs Tesla, Spacex, Starlink, Twitter, The Boring Company for YEARS in excellent fashion and very profitably is in reasonable - at least - health.
I think it's inconsistent to say that somebody can't claim Musk has mental health issues based on available information about his behavior, but you can say he doesn't, based on available information about his behavior.
If redSun was a psychologist it would be unprofessional, but I think we can reasonably understand they're speaking colloquially.
No, "available information" re Musk's corporate/innovative excellence and personal wealth are widely available and agreed upon via the financial press.
Re drugs and mental health, there is contradictory rumor and slander based on politics, ...
There’s definitely a lot of Edison aspects about him, between running and growing new technology industrial giants, and unnecessarily antagonizing many others and driving them away. (I don’t know if there’s anything that will be as much this anti-legacy of Musk as the Hollywood film industry was the result of Edison driving away all filmmaking from the New York area with his patent trolling.)
Ironic that Tesla is the century and a half old name that will end up being tied with Musk’s in future history.
I got a weird sense of deja vu reading this, I think dating back to a science fiction story probably 30 years old by this point, by an author that may well have been inspired by that WIRED article.
This bit especially: "Simply put a royalty on the links. If you want to reference a copyrighted New York Times article, then you’ve got to pay the author a little bit. And if someone else links to what you’ve written, then you get a small payout."
Except when I checked the series I was thinking of (https://en.wikipedia.org/wiki/My_Teacher_Is_an_Alien) it seems like they came out well before that article...so probably not. Either I'm thinking of a wholly different book or Bruce Coville is just a well connected guy, both seem possible.
After putting a few queries to ChatGPT (what would Nelson think of that I wonder), some other candidates are Earth by David Brin, or more in the YA line, Diane Duane: So You Want to Be a Wizard series, particularly High Wizardy since the manual comes in the form of a computer.
Probably burned a couple nickel's of compute to do the searching, although I doubt any portion of that would go back to the creators, alas Xanadu.
Bruce Coville wrote a trilogy of science fiction novels, "The AI Gang," about AI development that he published in 1986 and then rewrote and revised in 1995. I haven't read them since the Nineties, but they sound more like what you were looking for than "My Teacher is an Alien."
" There are no trails on the World Wide Web—instead, there are misattributed quotes, dead one-way links, constant plagiarism scandals, and widespread misinformation and mutual distrust. There are no trails on the World Wide Web—instead, there are misattributed quotes, dead one-way links, constant plagiarism scandals, and widespread misinformation and mutual distrust."
I'm not clear on how these aren't just two different ways of saying the same. The optimistic way, in which we have "trails" and a vague nebulous sense that they'll all be going something wonderful, worthy of a stately pleasure dome etc. And the pessimistic way. in which we have "links", many of which are dead and many of the live ones pointing to misinformation, distrust, plagiarism, and misattributed quotes. Why are the "trails" in Xanadu immune to the same forces that create all the crud in the current internet? If anything, from some of the description I'd expect them to be more vulnerable.
I'm getting the impression that "Xanadu" has only two significant advantages over the WWW. First, that it never managed to be implemented in messy reality, but could live only as dreams and promises, and second, that those dreams were frozen in place before Eternal September.
Indeed, the problem is people... and the complete lack of proposed technical defenses against the bad actors.
Writing an incorrect information is just as easy as writing a correct one, actually even easier because you don't have to waste your time fact-checking. The incorrect information could be corrected, but the correct information could just as easily be uncorrected... leading to edit wars? Ultimately the problem is that the words on the pages are unconnected to the facts on the ground, and anyone can assert any connection.
In a cooperative environment, "many eyes make all bugs shallow". In an adversarial environment, whatever method you use to correct the facts, the other side can mimic it to uncorrect them. For example, do you point out that a crackpot's page disagrees with hundreds of established sources? The crackpot will write hundreds of articles, and now your page disagrees with them. Soon there will be thousands of pages on homeopathy and quantum healing.
Then there are more subtle problems, such as adding a reference to a serious page, but misinterpreting its conclusions. Who is going to verify all such links? Also various ways of lying while being technically correct, that many journalists love so much. Cherry-picking. Etc.
It's thanks to the web being disconnected that we can have islands of relative sanity, such as ACX. Scott can set the rules about what is okay and what is not, without needing to get the approval of the rest of humanity.
Which reminds me: on web, you have not only static pages (with or without hyperlinks) but also web *applications*, such as discussions. And anyone can make their own. No mention of anything like this in Xanadu.
An insane story, and not in the oh-that's-cute way of figures like Emperor Norton. I guess it's good to know hallucinations were a problem with compsci long before LLMs? You can call it preconceived notions from growing up with the web (just barely old enough to have a before-and-after experience, one of first households in our town to get internet), but I think ending up with The Internet But For Schizophrenics would have been Bad, Acktually*. Are we really sure normal humans think in this sort of nonsystematic, freewheeling, trails-of-associations way? Are we really sure the blame for impersonal coldness and cruelty lies with the infrastructure that connects us, rather than PEBKAC? Never bet against PEBKAC! I just don't buy it. More limited claims, like social media specifically being poorly aligned, seem self-evidently true by now. I think you'd need to bring considerably more receipts to justify throwing out the baby with the bathwater though. Seems like one common sign of being Terminally Online is reaching the conclusion that "No, it's the *internet* that's wrong", yet not being able to log off it anyway.
Also, I regret following that youtube link for Suggested Reading Music. It made me distrust the moral aesthetic concepts (still think this is a weird Chomskyist conflation) behind the review even more, and this wasn't a particularly easy review to read through in the first place. Reversion to mean with previous years' contests: wild swings in review quality. I guess some things never change...
*although the pay-per-click/microroyalty thing would go a long way towards deterring spam and low-effort posting of all kinds, as well as elegantly solving the LLM plagiarism debacle, if it could be implemented
The details lie in the devil, for sure...you'd want the price low enough (zero?) not to incur Trivial Inconvenience penalties for prosocial things like building wikis, yet high enough to make the David Gerards of the world think twice. I'm fairly convinced that very low charges like a mere penny would disrupt the business model for many kinds of automated spam...more advanced implementations would also include a "refund the charge for mutually agreed upon positive interactions". But that's already starting to sound pretty complex for just a single site to manage, nevermind do so internet-wide by default.
(Of course, in an alternate world of "internet is powered mostly by subscriptions rather than ads", perhaps such payment infrastructure would already exist. But we didn't end up with that version of the web.)
I have heard that at some moment there was some kind of "web on mobile phones" in Japan (not html, but some special protocol). Users paid per minute spent online, because that was considered normal with phones. And about 10% of the money you paid went to the author of the web page you were browsing, which was easily measured by the phone companies. As a consequence of this model, there were many web pages in this protocol providing various services. I think I was told that at some moment, the number of web pages in this protocol (which didn't exist outside of Japan) exceeded the number of web pages on the internet.
Unfortunately, it seems to me that a model like this wouldn't work well today, because companies like Facebook and Xitter would optimize even harder to make the users never leave their web pages. The problem is, this model incentivizes you to create content that people want, but also disincentivizes you from linking someone else's content. When you are a small player, linking to others still kinda adds value to your pages (people are more likely to bookmark a page "links to 100 useful resources on some topic" rather than those pages), but when you are a big player, such as a newspaper or a social network, you can provide alternative links to your own content instead.
I already find this annoying when I read online newspapers that typically when they talk about some online thing (let's say, Khan Academy), the only thing missing in the article is the fucking link to the thing they are talking about. Instead, there are dozen links to other articles from the same newspaper on similar topics, or even articles from the same newspaper on other topics. Ironically, sometimes it is difficult to find a web service that many newspapers talk about, even by Google search, because those newspapers often have greater page rank.
Fascinating! Yes, it'd definitely be doable at the "hardware level"...not duplicable today because you'd end up with a bifurcated internet where mobile viewing effectively costs more, and good luck changing that equilibrium. People hate giving up "free". (Although maybe that's Working As Intended?)
Solving the linkage problem is hard, and that's one part of the review I'd have wanted to see fleshed out more...what exactly stops the sort of "Google summary sidebar" treatment (even before the AI Overviews), where you show relevant info snippets all within your own sandbox, and as a result that eliminates most traffic to the actual thing? Sure, you could charge for that snippet - but it'd be very weird to charge as much for that as actually visiting the page in toto. We see the same problem with Twitter going out of its way to throttle external links, especially to Substack. A link doesn't do any good if no one actually follows it!
'Are we really sure normal humans think in this sort of nonsystematic, freewheeling, trails-of-associations way?' -- I do think in this way. It's the kind of stream of consciousness you experience while out walking without headphones or anyone to talk to; or while doing random tasks around the house. I feel it used to be almost a default state of mind before mobile phones meant that I constantly have some source of entertainment by my side. It concerns me that when I look around now e.g. on public transport that not a single person is daydreaming, or (if I'm more critical) thinking.
I'm sympathetic to the idea that the Web as designed is techno centric rather than human centred - it's about facts and figures, not human experience. Certainly it could have been developed in a different way. On an entirely different note: the web was developed for good-faith communication between professionals - this is the fundamental reason internet security is so hard. If someone would design a system from scratch today, it would be totally different. Inertia is the reason cyberattacks are so common.
I guess this is a Typical Mind thing...I have a separate subprocess that flits around associatively, usually choosing to play whatever music suits the mood, but is also responsible for things like reflexive pun generation. For Real Thinking(tm) though, that part gets sidelined, and we bring up Systematic Filing And Calculation. One could certainly say that's been influenced by the web - even though I don't do much with phones, obviously I'm online, and have never been too keen on nature retreat type stuff. (Not because of lack of internet, but because I just like lightbulbs, heat, running water, lack of bugs, etc...) So it's hard to do a true controlled experiment, see how my thinking would change if durably distanced from our cybernetic minds.
It is indeed depressing that most people in public seem hopelessly addicted to their phones. Not just on public transit, but walking the streets, or driving, or in the middle of crosswalks (and then they have the audacity to get mad when *they* bump into *you*), or standing by awkwardly and helplessly while you bag their groceries. Such a lack of being present in Simulacra Level 0, of potential human connections never made. I don't really wanna judge it as "not thinking", but...a lot of it certainly is passive consumption. However, the fact that it's apparently so addictive is part of what makes me question the Xanadu narrative - somehow this supposedly deeply inhuman internet manages to ensnare billions of people anyway. If it were really so alienating, many more of those people would be daydreaming, noticing the world around them (even doing this while listening to music is an improvement), seeing those of us "open people" on the bus wondering where all the thinkers went. Instead they while away their leisure watching Chinese spyware propaganda and stoking division for their fellow man...
This was a really interesting review and I can't help thinking I'd enjoy a biopic. A family member has worked in digital records management and some of the principles seem to cross over. If a public body is the potential target of a Freedom of Information request, it's in their interest to make files as accessible as possible and as quickly as possible - which means good naming conventions, version control, and metadata. Cue moaning from staff who want to have a file called Misc or Random.
In a round-about way, modern systems with cross-links, especially shared notion pages and vast obsidian pages with peoples' dozen hyperlink plugins, *are* a lot closer to Xanadu than the normal web.
The vision of this project sounds to me like a collabrative Zettelkasten, which is an idea that pops into my mind every so often.
Perhaps it won't ever work on the public web, but I do think that various groups and organizations can benefit from having their own intra-memeplex, in a similar way to the commonly found intra-nets.
I recall that Phillip Hallam-Baker once proposed "The reason the Web works and previous attempts to implement [Ted Nelson's Project] Xanadu failed is that the Web has 404 Not Found. Ted Nelson insisted on referential integrity and since achieving that is a byzantine generals problem, the systems never worked." Certainly the Web has a great advantage in that any cretin can do a marginally-adequate deployment of a web server.
Intriguing article! I think I see Vanevar Bush's vision, but I'm not sure I can see Xanadu's vision.
In one part you say:
"The three quickly figured out a new system that would allow users to reference and link to specific parts of a file—they called these links tumblers, and made them work with transfinite numbers."
What does this mean? I'm very familiar with transfinite numbers --- but I don't understand how it could relate to linking to specific parts of a file, at all. Anyone have any idea?
On my first read through your comment, I thought you were referring to a TED Talk, and was momentarily thrown for a loop. I may start referring to them as "Talks of TED."
If you're interested in modern ideas along the lines of the memex and Engelbart's "Mother of All Demos", I recommend Ink & Switch's essays: https://www.inkandswitch.com/
(The essays may be interesting to a general audience, but they are mostly aimed at software engineers, since Ink & Switch is very interested in how to actually build these things - including many prototypes and occasional spin-off products.)
Nigel Calder once posted this about a techno-prophecy by Maurice Wilkes in 1964 of a world information system on networked computers. Wilkes made no suggestions about how it might be implemented, certainly no mouse or hyperlinks, but Engelbart was not the only one keeping Vannevar Bush's vision alive.
And then we tried again and called it Arbital. Admittedly, Arbital worked pretty well, as-implemented, for an _extremely_ narrow use-case, which puts it well ahead of Xanadu.
Weirdly, this reminds me somewhat of TV Tropes, and how pages are linked there. Over at TVT, there is a large but finite list of tropes one can use to list the ideas that two works have in common, which sounds like what Xanadu/Mimex was trying to accomplish. Two pages on two different works list the tropes that are associated with each work, and viola, you have a set of ideas connecting those two works. Click on one of the listed tropes, and you get a list of other works that utilize that trope. TV Tropes is, therefore, a hyperlinked semantic network.
Not sure how that would apply to Xanadu (or the web), if it could be. But here's another potential weakness: TVT is notorious for being a rabbit hole. Click on any page, and several hours later you suddenly wake up, head full of the works and tropes you've been reading. Xanadu would be worse.
That’s an interesting point! I had long noticed that TVTropes was the one website that more reliably produces multihour binges than Wikipedia, but hadn’t quite understood why. If it’s something about the siren song of the infinitely recursive cross-linking that works both ways, that would make sense.
Yeah, I think it is. It helps you discover new works that have something in common with a work you already know, around an idea (trope) that you didn't even know was a thing. That sounds to me pretty close to the vision they had for this Xanadu thing.
The hook of the Memex is strong - that's a machine anyone who works with information would dream of having.
I think the review is built around how sad we should be that we got the WWW instead of the Memex ... but this seems like a false model.
The Memex is described as a personal knowledge management system - and in the last 20 years software tools (from Org Mode to Obsidian) have become pretty common.
I'd argue that Obsidian more or less fulfills all of the qualities that Bush described - maybe with needing to take the extra step of explicitly creating files to reference for each individual piece of work - article, book, video - that you consume. (I would argue this constraint is implicit in the description of the Memex since Bush describes the machine as physically containing copies of everything you've read).
So I'm not that sad Xanadu never came to book - or that we got the WWW instead.
My model
- The dream of Xanadu or the Memex makes a lot of sense for private information management - where bidirectional links and margin notes are super valuable. We have this today with Obsidian.
- The WWW makes a lot more sense for no trust global communications / content distribution networks. Xanadu would not avoid the problems with content on the WWW this review describes.
As someone who has certainly heard of Xanadu before (I remember first from a DEF CON video in particular) and got quite curious at least from the historical side, the author here has fine taste in my opinion. it should be noted that Ted Nelson actually had a YouTube channel for anyone interested, seemingly a lot of content in relation to this: https://www.youtube.com/@TheTedNelson/videos
From there in particular, there was a series of "Xanadu Basics" videos that explain quite in detail his actual thought processes for the system and how it generally functions, that's generally one of the only organized attempts I know of from them: https://www.youtube.com/playlist?list=PL0pAM29gc1g9o58LWnoXkqAVWq8y335hX
Whether or not that's worth anyone's time here probably varies, I once again simply see this as quite interesting from a historical standpoint and nonetheless also think that, despite to what extent his ideas may have been considered redundant nowadays, the sheer quantity of his time at least with the topic of organizing information seemed promising to me. Quite nice to see this post pop up notification wise as a result.
I'm one of the many nerds who bought Nelson's books in the 80s and pored over them. I got to see Engelbart demo his system live just before Stanford turned off the one computer it ran on, mid 1980s. That's a treasured memory.
The math that makes neural networks possible was published in 1980. It wasn't until the 2010s that the gear was fast enough to run them at scale. People forget how mind-bendingly slow computers were before the mid-90s. Neural networks are what make associative computing feasible.
But! The one Big Idea missing from the early visionary days was Digital Money.
There's a lovely low-res game called "Kentucky Route Zero", an immersive journey through the Underworld of Appalachia. One of the sequences is using a buried computer mainframe which runs game-within-the-game: an unholy mix of Xanadu and the old Adventure game. KRZ is a great game to watch playthroughs of, you don't have to actually play it.
Computers were both slow and expensive, so even the stuff that Engelbart was doing (which did not require neural networks) wasn’t very practical from a cost perspective at the time.
Can someone explain to me what the difference would be between Xanadu and Wikipedia? I imagine decentralization across servers is one, but I'm looking for differences regarding the philosophy of hyperlinks, version control, and so on.
Copyright seems to be the biggest sticking point. Wikipedia is powered by the free/open sentiment, and with anything like a pay-per-use model it would've been a non-starter.
This is really interesting and fills in a lot of gaps of my knowledge. I studied under Andy van Dam and he often tossed around a lot of these names—nice to have the blanks filled in about how they all participated in the history of the Internet.
Tbh, I was surprised this article didn't address Urbit (https://urbit.org/) —since that seems like the modern attempt to completely rewrite the Internet into something better.
I really wish this went into the rise of Roam, Obsidian, and similar, because you can kind of see an in-between of Project Xanadu and the World Wide Web, especially in the use of back links
I'm a bit grossed out how people *want* the founders of the web they use to be Proper and British and Adult and Correct (and Male).
Idk about you, but I want a web founded by adhd ppl, by crazy kinksters, by trans folks, by nerds and enthuisasts and ... you know? people who have something to hide. They'll keep things save.
And, at the same time, people who don't hide. Because they won't be compromised.
Most Correct People are incorrect people in hiding, and I don't wanna live in their world.
So, do you like Xitter? It's certainly much crazier and less Proper and Correct than pre-takeover, and yet I have a suspicion that it's not your cup of tea.
I want to live in a world where people don't need to self-censor.
As far as I can tell, these days twitter is a place where nazi hate speech is very much permitted.
Thus, it's a place where vulnerable people have to self-censor heavily, or they get publicly abused otherwise.
Or in other words, it's not a place "to be improper", it's a place where you can openly be a nazi, or act in a way such that nobody can prove you aren't a nazi. The latter is what "being proper" often means.
And this is why I hate the "respactable ideas have to come from respactable people" way of thinking: It usually forces people to act like nazis-in-hiding.
Which means no non-nazi can tell apart their supporters and their enemies. And this gives hateful ideas more power over everybody's lifes. I very much want to live in a world where this cycle is broken.
And I start with refusing the idea that every non-cis-straight-mono-vanilla-english-christian-conservative-wealthy-binary-man has to do their utmost to hide that they are different.
Tolerance is a social contract. I don't tolerate ideas that don't tolerate me existing.
It's funny how your side pretty much controls the entire internet and still thinks that its persecuted, underdog bias is a hell of a drug. You think that people who disagree with you don't have to self-censor in order to avoid being branded a nazi?
Scolds have stopped being exemplified by The Correct And Proper British White Male a few decades ago, they're now the sort of people who use expressions like "ideas that don't tolerate me existing", but somehow it doesn't seem like we've come any closer to a world where people don't need to self-censor.
If Xanadu is so awesome, I expect it would see use in a few private companies, like a juiced-up Confluence. But I expect most private companies will find that the costs of maintaining a private docuverse just for them is too high for just a moderate benefit.
Is Xanadu one of those technologies that absolutely needs to have wide adoption and network effects before philistines like me can see its value?
The concluding sections of the essay are comparing existing software to vaporware. It’s not hard to make the latter sound better than the former.
Selection by association rather than indexing sounds profound, but it’s there in HTML, which allows links everywhere and doesn’t force a top down directory structure. Bidirectional links? Not built into HTML but Google implemented a “find all pages linking to this web page” search. The “lame text format and a lot of connected directories” criticism applies more to Gopher (which predated the World Wide Web and was outcompeted by it).
“See every quotation in its original context” requires some sort of enforcement mechanism to prevent people from quoting without providing context. “Quote without limit without permission” with an automatic royalty payment requires changes to copyright law, which would require renegotiating (or abrogating) international agreements. In an alternative timeline where Nelson actually managed to get a version of Xanadu built and released, he probably would have done it by placing these goals in a “to be addressed in future releases” bucket, where they might still be today.
I loved, loved, loved reading this. I have both the Mother of All Demos and the student film saved for later, and am considering getting the autobiography. I can only hope these things are half as interesting as the reviewer's descriptions of them.
Someone once wrote "The Web is wonderful and horrible -- it's like the really messy desk of a brilliant but deranged professor -- but who has a dim-witted graduate student who has more-or-less indexed everything. If you rummage around, you can find nearly anything."
What we *want* is Xanadu, that is, the Web, but everything properly organized, indexed, linked together, annotated about reliability, etc., etc.
What we *need* is a *smart* research assistant to organize everything on the Web. And it seems to me that it might be *just* within the foreseeable capabilities of an LLM-driven automaton to be that research assistant. For instance, if you want to judge the reliability of something, you need to compare it with similar things with known reliability, check how it's reviewed by sources with known reliability, etc., etc. This is what people do, but it seems within possibility that it could be automated.
Or perhaps, the Xanadu database itself isn't materialized but whenever you look at something, the automaton updates all the associated links. "Just-in-time Xanadu annotations".
That's a smart idea; apply some associative Xanadu style system to what already exists. The LLM analogy is suited; they seem to act much more like brains.
I thought this was overall an excellent overview of some important history, if a little nastier than it needed to be.
One correction: You say "*Computer Lib* was his lamentation over the industry’s disdain for hypertext, and *Dream Machines* was Xanadu’s manifesto." I don't think that's an accurate summary, of either section.
*Computer Lib (You can and must understand computers NOW)* was intended as an overview of computer technology for non-technologists, including sections on programming languages, data structures, machine architecture, early hacker groups, and much more. The format and attitude was similar to the Whole Earth Catalog: look at all this cool stuff. Some strong opinions and even some lamentations, but that was not the dominant message.
*Dream Machines (New Freedoms Through Computer Screens)* was about the research frontier of interfaces and visions for human/computer interaction, of which Xanadu was just one example, It included eg very early CGI, real-time image processing, and ELIZA. It was not that different in form from the other section, just more forward looking.
I am not very objective about this book which had an enormous influence on me. I encountered it when it first came out, in 1975 or so. This was before the first personal computers were available and before there was any real mainstream idea that normal people should be interacting with computers at all. Both sections conveyed an irresistible air of excitement and revolution, and the idea that technology might actually be wrested from its military/corporate origins and put to humanistic uses. So, yes Ted Nelson may have some character issues and be not very good as a system builder, but he deserves enormous credit for his ideas and writing.
Thanks for this excellent reminder of computer history and how the a simpler approach like the Web can outperform a more kitchen sink approach like Xanadu. But you have to love Ted Nelson for how he really tried to get the average person into computers in an era where big iron dominated computing and computers were intimidating. Everyone needs to glance at Computer Lib and get a sense of how computing was in the 1970s and be really really glad to be living in a time of computers everwhere!
If you want to try note taking/word processing like this describes, try obsidian editor. It has a learning curve, but it's super powerful, and really nice to work with, and you don't need to know how to do everything to be able to use it well. It lets you super efficiently link to files (you can also link to sections of files) or create a new file as you type in a current one. I'm tired and not selling it well but if you read the idea of the memex and thought it would be cool to try, try obsidian or look up videos of people using it
The comments on ddos attacks are well considered and we did think about all that, and when I have a better solution for it I'll implement the system with it, for now I'm considering about reviving it. Roger Gregory
At least to me it is clear that "your review" means "a review by one of this substack's readers" - otherwise Scott would have written "thy review".
Well it's not addressed to you then.
I agree "Review Contest:" would be a clearer prefix for the title.
"One of all y'all's review"
Ironically, were we doing currently on a Xanadu, this would not be an issue...
It's 1965. You're watching your favorite television talk show. They've just started broadcasting in color, and you've got a color television set, so you're pretty excited. At the appointed hour, the jazzy theme song plays, and the host comes out and begins to deliver his monologue. A few topical jokes at first, which aren't your favorite, but they're amusing enough: a joke about LBJ's twangy accent, a bit about women in miniskirts driving on the Interstate Highway system while listening to the Beatles, a pitch-perfect imitation of United Auto Workers president Walter Reuther ordering dinner at a fancy French restaurant. Following some banter with the bandleader, the host sits down at his desk and picks up some envelopes. "And now," he announces with an impish grin, "we come to the part of the show where I read your letters."
You haven't sent any letters to this talk show host. Are you confused?
Hoooookay, buddy
Hoooookay, buddy
I didn't see the two comments that replied to yours but I'm interested in what they said. The OP deletion that didn't reply to you is pretty easy to guess but it sounds like the two replies to you were maybe pretty spicy so I'd love to hear them if you remember, or at least a paraphase if they can't be printed here
This is a cool history story and I liked the writing. But the review sort of lost me in Part Eight, where it showed me a screenshot of a Xanadu browser, and it asked: "Why shouldn’t the internet look (and work) a little more like this?"
It's possible that this is the sort of thing that you have to use for ten minutes to understand. But it looked to me like a normal (image-less) browser with a bunch of tiny unreadable slices of text on either side, and I don't feel any particular need for the internet to look more like that.
I would agree with the author that there are a lot of problems with the internet, but I'm left with no idea how Xanadu thinks it's going to solve them.
Concretely: Xanadu offers "connection, accountability, verifiability", but it doesn't explain how it's going to get any of those things.
Verifiability: Is this going to be one of those Internet things where you have to upload a photo of your drivers license before you can post? If so, does it offer protection against getting doxxed? If not, how is anything verifiable?
Accountability: If you post something deliberately wrong on Xanadu, what specifically happens?
Connection: Can a spammer create a page about their new memecoin, and "connect" it to the most-used page on the Internet so that everyone in the world sees it? If not, why not?
There are a lot of people that want connection, accountability, and verifiability. But it's not enough to handwave saying "this project solves those problems". These are actually-hard problems and you have to explain how you solve them.
I have nothing to add to this excellent point, and want to just "like" this comment, but ACX doesn't enable the like button, so you get this reply.
Seconded!
There is an extension "ACX Tweaks" on the Chrome Extension store that modifies the styling on this site in a number of user-configurable ways, including "Show hearts". There may be similar ones for other browsers. Enough people are using this that I can tell the two parent comments have 17 and 15 hearts respectively, and yours has three.
I can see them on the mobile app :)
This. The whole project strikes me as something which has to be proprietary and controlled by someone. The web we have is decentralised, because it basically lets anyone who can rent server space put whatever they want in a window of whomever’s willing to open it. Xanadu seems much more restrictive; you could build Xanadu and stick it on the web, but I don’t think you could build the web and stick it on Xanadu.
(Technically, there is also some centralization on the web, for example until a decade ago the 13 root DNS servers were all controlled by ICANN.)
Yes, Xanadu would either need a centralized database (like Wikipedia) or a way to make multiple databases cooperate (like Fediverse), which opens another bag of problems.
If I am following the description (no guarantees there) Xanadu could be the web. I think it works like this: every time someone wishes to borrow an idea from one piece of media, a link is created back on that original piece of media with a link to everyone who has borrowed anything from it. Simultaneously, there is a link back from the new article to all the originals it borrows ideas from (I don't know how you stop people from claiming ideas they didn't originate, but the current web can't do that either). In theory, one could read an article, track back to where it got an idea, see everyone else who borrowed an idea from that, and go merrily on. It's network of ideas, not just documents.
I don't use pubmed. Can you briefly describe how it works, and what you see as the parallel to Xanadu?
I don’t know pubmed in particular, but many academic paper databases now have a page for each paper together with links to click to see all papers cited in this one, and all papers citing this one. The latter isn’t always complete of course - at best it’s all papers published in journals and volumes indexed by this site. But that’s often pretty good.
Here’s one of my papers at Google Scholar: https://scholar.google.com/citations?view_op=view_citation&hl=en&user=4AzuGlEAAAAJ&citation_for_view=4AzuGlEAAAAJ:ufrVoPGSRksC
And at PhilPapers: https://philpapers.org/rec/FITACA-3
They show somewhat different numbers of citations (175 vs 110) but I can’t tell how many are missed citations at one or duplicates at the other.
This is the sort of mechanism I advocate to grad students who want to learn some literature, and ensure that their response to a paper is somewhat original.
What’s to stop them not putting the link in though? Or is this just the web but with very strong norms of putting links in (and where links are somehow reciprocal)?
I assume the process is automated somehow. That is, if you copy something off someone's document, the system knows this and creates the link back to yours.
Well, but you ignore most of these links to you. Otherwise spammers will have a field day.
So a sufficiently popular piece of original media will be seen as 5% original content, 95% links to everyone who ever riffed on that content? And if the original is sufficiently controversial, I suspect half of those links will have labels that are cheap little attacks against the original with the promise of more to come if you click on the link. Outrage galore, and DDOS to useful communication.
Yeah, it's not clear to me from the description where the tracks are located, but I assume it's all on the servers somewhere. This is supposed to be a distributed computation model, so it could be anywhere, I suppose. Were we doing this now, some sort of voting system might be put in place, like in Wikipedia.
The cypherpunks of that era had some thoughts on deterring spammers from bringing down their anonymous email/remail systems. Hal Finney's Reusable Proof of Work was intended for that. Now we have Bitcoin and other crypto systems using distributed blockchains, but spam is handled by AIs looking for suspicious indicators rather than requiring that any user pay or verify somehow that they aren't a bot.
Though our AI systems could use proof-of-work in the email as one indicator in favour of a message not being spam.
Yep, completely agree. This Xanadu thing seems like a really lame how-to-do-more-of-the-same-just-in-a-different-medium solution. It reduces the time and effort for a scientist who reads a paper, notes an interesting reference in another journal, and then goes to read that journal. Or the newshound who reads a piece of news, spots an interesting reference to a piece of history she doesn’t know, and looks it up in an encyclopedia. It solves those specific problems, but it doesn’t do two things that it claims to do:
1) It doesn’t really reflect the “variety of concepts and concept structures in a way that makes them maximally available and useful to the needs of the human's mental-structure development”. People make all sorts of weird connections that cannot be represented by this very dry source-of-quotation structure. When I read papers, I think of poems; when I read poems, I wonder about crystal structures. The weirdness of the existing web actually reflects the weirdness of people’s brains much better.
2) It fails to respect the fact that the web really is a different medium. It’s just an attempt to take the professionally-written journal and encyclopedia and digitise it. But the web really allows anyone to write, and accepts the fact that most of us are terrible writers. Terrible writers and videographers need very loose structures to contain their blatherings, and we got it.
> Can a spammer create a page about their new memecoin, and "connect" it to the most-used page on the Internet so that everyone in the world sees it? If not, why not?
Yeah, that was my first thought after reading the proposal of bidirectional hyperlinks. They are great if both sides agree that the two documents should be linked. It is a problem when one side wants to make a link, and the other side does not want to make a reciprocal link. Current web solves it by links being one-directional, so everyone decides what their document links to.
Even assuming good faith, some documents are just vastly more important than others. Imagine millions of documents linking the definition of atom in Wikipedia. Should the Wikipedia article on atom link all of them? (At the very least, we would need a system to organize the outgoing links, maybe group them in categories such as "textbooks referencing 'atom'" and "personal web pages referencing 'atom'" etc.)
But maybe this is too much to expect from version 1.0. The e-mail protocol worked well at the beginning, and remains kinda usable today, despite being completely unprotected against spam. Maybe if bidirectional links became the norm, we would figure out some solutions to abuse afterwards.
One possible solution would be that the bidirectional hyperlinks need to get the target's approval, and in the meanwhile act like unidirectional hyperlinks. A mechanism similar to "friend requests" on Facebook would show you "hyperlink requests" and give you the choice to approve or ignore. (Of course, there is still potential for abuse, for example the other party might change the contents of their page after you have approved of the link.)
So, maybe something like this could start in a high-trust environment, and gradually adapt to less and less trustworthy users?
>Even assuming good faith, some documents are just vastly more important than others. Imagine millions of documents linking the definition of atom in Wikipedia. Should the Wikipedia article on atom link all of them? (At the very least, we would need a system to organize the outgoing links, maybe group them in categories such as "textbooks referencing 'atom'" and "personal web pages referencing 'atom'" etc.)
Ouch! Excellent point! My immediate worry was about the analog of distributed denial of service attacks, but, even with good faith actors, just _organizing_ the backlinks is, as you said, a problem.
>One possible solution would be that the bidirectional hyperlinks need to get the target's approval, and in the meanwhile act like unidirectional hyperlinks. A mechanism similar to "friend requests" on Facebook would show you "hyperlink requests" and give you the choice to approve or ignore.
Hmm... To first order, yes that gets around the DDOS analog. To second order, though, in addition to the problem that you already articulated, there is the problem of now needing a (human?) decision for each of those millions of links to popular pages. _Maybe_ that could be automated with today's AI, but I suspect that there is a broad set of vulnerabilities...
That sounds like the problem Bush was trying to solve. You have a vast mass of documents, how do you organize them? Well, so far as I can tell, the user creates a file that organizes them in some way useful to that person. Again, so far as I can tell, none of these guys gave a thought to anyone *sharing an organizing framework* When you look at someone else's file, all you see is your connection to it. But how do you get to the next file/book/idea? There doesn't seem to be a system for that, unless I missed it.
We have Google.
Until something better comes along (LLM's ain't it, at least not yet), keyword search is what we have.
Seconded! In addition to your concerns about whether there is any way to actually implement verifiability and accountability I'd add:
If every link creates a backlink (stored on the linked page????) don't we automatically get a version of distributed denial of service attacks that chews up massive storage on the target page?
I think that, in our actual time line, the invention and deployment of search engines does most of what was suggested by Xanadu and lacking from just the linking technology of the world wide web - yet reasonably feasible.
In general, I think that this review (and, frankly, the architecture of the world wide web itself) takes too little account of what to do about malicious actors.
Orthogonal to that, but still an unsolved problem, even in the _absence_ of malicious actors, I think the promise of links (even very sophisticated ones) has been oversold as improving intellectual discourse more than they really can. (ignoring AI for simplicity) A rich set of links plus search capabilities gives, roughly, an exponentially increasing number of documents with search depth - but human bandwidth doesn't go up. Most arguments are, of necessity, unanswered, most links untraversed, most ideas ignored.
My interpretation (IANAP) is that with a Xanadu system you see the source of every piece of text automatically. Nothing can be misquoted, and manual attribution is unnecessary. I cannot tell if this is either practical or desirable, but it's interesting.
Some of these ideas may be workable if they were combined with Joe Armstrong’s ideas on changing the web. See: https://joearms.github.io/published/2015-03-12-The_web_of_names.html
I actually think there are some reasonable answers to all of those points given *just* the hard-linking! Having not used the software but to take a charitable view of what was presented in the review:
Verifiability: You take verifiability to mean verification of identity and (possibly?) assume that also helps some verification of claims. I think Xanadu style linking deals more directly with the latter. If I claim 'Bob wrote X; how dare he' then within a Xanadu style system presumably I would be expected to link to verify my claim or whoever reading would reduce their credence in that claim. Of course, I can make greenfield claims without linkage "X causes Y but you won't hear that from Z because they're lying to you" but clearly we are still in a better situation than where we are now where 'Z' is often misquoted and framed without easy enough verification for the average reader to check in the middle of reading.
Accountability: Don't you think that if every deliberately wrong post had a list of indexed links to every debunk article accessible on the same page of your browser you would have a harder time convincing people of deliberately misleading things? Wouldn't you lose credibility otherwise? I'd say that is a system of accountability. Again, not saying it's perfect, but seems like a clear argument could be made in favor of it being better than what we have now and that's the bar to pass. Kinda, in a way, a twitter notes for the whole web.
Connection: Yes, a spammer could reference a popular page and get linked to it just like a spammer can currently reference a popular topic to show up in search results on that topic. Presumably the linking system would employ some type of page rank so that more pertinent links float to the top of the associated link list over the crypto scams.
> Accountability: Don't you think that if every deliberately wrong post had a list of indexed links to every debunk article accessible on the same page of your browser you would have a harder time convincing people of deliberately misleading things? Wouldn't you lose credibility otherwise?
No, not at all. It's already the case that many of the posts I read online are debunked by other posts I read online. This looks like anything from the several ACX posts where Scott responds point-by-point to a debunk article by someone else about an earlier post of Scott's, to some random person retweeting something I've never seen before purely so they could tell the world that the author of the original tweet is a Nazi or a retard or a retarded Nazi. The existence of some post B purporting to debunk some other post A gives me basically zero information in and of itself about whether post A, post B, both post A and post B, or neither post A nor post B are correct/interesting/worth my time to read/etc.
Encoding responses to posts into the system that displays posts to a reader doesn't seem drastically more innovative than e.g. the tumblr reblog system. It certainly doesn't seem like it would meaningfully change the dynamic that anyone posting on the internet about anything could be an insightful genius, a dumbass, or a spammer; and there's no way to determine that without actually reading their post critically and thinking about it.
I rather suspect that illegibility to the proles is a part of the point - the fantasy of some return to pre-Eternal-September Internet, not that most of the people talking about Eternal September would even be able to have any personal experience of pre-Eternal-September times (I certainly don't!)
To me it looked pretty intuitive. The current model of writing a webpage is (roughly) that you of sit down in front of a blank screen and typing your text document, when it's finished you hit publish and it's available for all eternity. The analogy is to writing a letter on a sheet of paper.
In this editor, you have multiple documents open and you can copy from them into what you are working on. The editor tracks this (those are the tiny texts on either side) and presumably if those are edited then so to would your document. There's no issue with citations - the format itself tracks where things come from, and the relations between them are as important as the content itself.
Maybe it helps that I've been reading postmodern fiction? Flann O'Brien borrow characters from other works because he claimed there were enough fictional characters already in existence; and one of his novels contains adverts and translations of poems as insertions. Borges played the same games with authorship and citation and etc.
I'd have liked to see an analogy in the review between the classical view that information in the brain is (somehow) stored in brain cells to the modern view that it's stored in the connections between neurons. The connections between texts being at least as important as the texts themselves seems to me the underlying idea of Bush and of Nelson.
My reading of the review is:
1. Guy has cool conceptual idea but no real technical knowledge about how to implement it
2. Lots of folks with actual technical knowledge try to make it happen on and off across multiple decades
3. They fail
4. Isn't it tragic that we somehow don't have this thing that would be so awesome?
To me, (4) should be "So obviously this thing had fundamental underlying technical issues that made it unworkable". But that idea wasn't really addressed in the review, which I find mildly disappointing.
(I did enjoy the read. It's an interesting story and history!)
Honestly that's even too charitable. My takeaway is "a part for giving people lots of warm fuzzy feeling about the power of liberating information and so on, there is no actual concept there. It's just a guy who thought how awesome it would be to have awesomeness, and somehow his charisma and eloquence made some otherwise intelligent people follow him in his quest to find the phoenix he could not even properly sketch"
It sounds to me sort of like a gigantic, open version of the old Apple hypercard system. Which was great for its time, and still sort of cool in an artistic way- you can throw together stacks of media, and hope for creativ inspiration to strike, like a non linear powerpoint presentation. Sort of like a software version of a hedge maze or Chinese Scholar's Garden, the nonlinearity is the point.
But then, those only work because they're closed off. I have no idea how you'd make a version open to all publishing like the WWW. But maybe it could work on a much smaller scale, with central control, sort of like a museum that's open to all visitors but only the central staff can publish there. And then the visitors can wander around at random, having their own unique experience, without the spam and junk of the general WWW.
Tim Berners-Lee went on to propose and advocate the so-called Semantic Web, using RDF and OWL, which somewhat relates to Nelson's ideas. Although RDF and OWL are sometimes used for niche applications, it's safe to say that the Semantic Web never took off, in spite of full backing from the World Wide Web Consortium (W3C).
Perhaps the real lesson here is the bitter lesson. All of these rich hypertext approaches, whether Bush's, Nelson's, or Berners-Lee's, rely on sophisticated formalisms that require a degree of human effort at write-time. (OWL, for example, was derived from knowledge representation formalisms used in symbolic GOFAI.) Maybe that's just a mistake. Maybe the way to go is with general learning algorithms that compress knowledge into models. Maybe GPT-5 is better than what a fully developed Xanadu ever would have given us.
I think the problem may be harder than that. As Bertrand Russel said somewhere, acts of thought are like cavalry charges in battle, they're expensive and you can't do many of them. A device that makes analyzing arguments, understanding their history and all of their connections ... are pretty much like good college courses, they take a lot of work, they aren't quick, there's not that much demand for them, and 99% of the time, you don't need them.
I mean, the panic-du-jour of the Internet is that too, too many people type a question into Google and take Gemini's two-sentence answer at face value. And because they don't even bother to click the link that Gemini provides (to verify that Gemini didn't screw it up), whoever is getting click-paid for the page Gemini extracted the answer from doesn't make any money off it.
Or I could phrase it that Xanadu is *great* for people who routinely work at the highest levels of cognitive complexity. But most people never do that, very few people are paid for it, and a lot of people aren't capable of it. In the end, *What is the target market for Xanadu?*
A stunted version of the Semantic Web has been implemented, and it is taking off like crazy. It's called the "Model Control Protocol", and it is basically how some software system describes itself to an LLM so that the LLM does not hallucinate nonsense when generating content for that software system. Example: the MCP for Python allows an LLM to generate code that is syntactically correct Python.
MCP is great for extending the capabilities of LLMs, but I don't think it has much to do with the Semantic Web (as based on RDF and OWL).
I was wondering if that was the point the piece itself was reaching towards. Particularly with the idea that human thought is fundamentally associative.
I think Gwern has some insightful commentary on why Xanadu failed: https://gwern.net/xanadu I would summarize it as:
1. Most of what it's trying to do, beyond what we got, just isn't that useful,
2. Its ideas about how to handle copyright were unworkable,
3. A public version would have had serious problems with spam and abuse.
Really, I'm not sure I buy this idea that this was all somehow contingent on Engelbart being discredited -- is it even really true that Engelbart was discredited? That's certainly not what I'd gathered from what I'd read elsewhere... it looks to me like there's so many reasons for its failure (the reasons Gwern mentions, the fact that the people making it didn't actually have a clear idea what they were making) that I'm not sure things could have gone hugely differently here.
The WWW succeeded because it was simpler than Xanadu. Worse is better: https://www.dreamsongs.com/RiseOfWorseIsBetter.html
Gwern explicitly writes: "Xanadu wasn’t the victim of 'Worse is Better'; it was just a solution in search of a problem." And makes a number of arguments to this effect. Worse-is-better isn't an iron law of nature; C++ and Java have had plenty of market success.
Worse-is-better might not have been why Xanadu failed, but it was why WWW succeeded.
To some programmers, the success of any language that isn't a LISP dialect just serves as confirmation for Worse is Better :)
To some other programmers, it is the success of any language that isn't Smalltalk. :)
I was very confused by the linked page, which seems to expect the reader to have a high level of context about programming language and operating system design.
I eventually figured it out with the help of this other page: https://www.dreamsongs.com/WorseIsBetter.html
Quote: "The concept known as "worse is better" holds that in software making (and perhaps in other arenas as well) it is better to start with a minimal creation and grow it as needed."
Yeah, in retrospect that would've been a better url to give.
Wow, this is great, thanks for linking!
One question I have: Gwern writes, "Like Douglas Engelbart, Nelson enjoys (if that is the word) the honor of being a living embodiment of Cunningham’s Law: becoming one of those historic figures whose importance was that they were *so* wrong on *such* an important topic they helped popularize that they inspired others to become right."
Do you have any idea what he's talking about re: Engelbart? Is there an important respect in which the Mother of All Demos was conceptually flawed? I vaguely recall hearing allegations that Engelbart had some unworkable-in-retrospect Bush-esque ideas about how it should be used, but I can't find details and the demo looks fairly similar to how mainstream computer UIs ended up actually working (at least once Google Docs made multi-user editing a thing).
Adding my thanks for linking this - very interesting. I am curious that Gwern doesn't discuss or even mention scientific research as a use case; if I've understood this review correctly, that was the original motivation for the whole thing.
(Much) more speculatively, perhaps Xanadu feels more useful to people who process visual information in an unusual way?
I disagree with these general points (although admittedly I have also dreamed up ideas similar enough that people pointed me at Xanadu):
1. I think there's lots of information that could be more useful with a bit more than just text. It's really common when researching controversial stuff, particularly in politics, to have to follow claims through lengthy chains of links only to eventually figure out some part of it was bullsh** or it was all some telephone game. I'm convinced we can do better.
2. Lots of stuff isn't copyrighted and that Internet does fine with a wide mixture of stuff with different licenses.
3. This is a problem, but also it's a problem with tons of existing systems. I think a richer system is going to be no worse than social media (looow bar, I know).
[EDIT:] Reading through Gwern's writeup I see I misunderstood the second point, Xanadu was trying too hard to SUPPORT copyright. This is a reasonable critique. I also would say I think Gwern makes the case that in practice the UI they built for Xanadu was kind of terrible, which is a real death knell if you want to get everybody to use it.
>Xanadu was trying too hard to SUPPORT copyright.
I'm writing out of ignorance here, but one question that I don't know the answer to about Xanadu was: Is there an easy way for the user to know the cost of a payment to a copyright holder _before_ they access the copyright holder's content? One thing I find irritating about substack is that it isn't clear how much a subscription will cost before subscribing (though I may have overlooked this - but not for lack of searching for it). This can make or break a payment scheme.
At least for Substack, you'll know partway through the process but before you actually pay anything.
No idea about Xanadu.
Many Thanks!
As an (advanced) end-user lived and worked through the computer revolution from 1977 until the World Wide Web revolutionized the Internet (ca. 1996), I don't remember that Engelbart was either discredited or marginalized. As best I can remember, he was respected in the personal computer computer community (even though he didn't like the idea of standalone personal computers) and then the Internet community. I remember lots of articles in the popular computer magazines praising him as a visionary — especially when the graphical interfaces of Lisa and then Macintosh started blowing people's minds. Englebart was credited with the vision that created the first GUIs.
Agreed! I've always heard Engelbart praised, not disparaged.
Nobody knew about the EST nonsense; if they did, it was just some goofy thing a lot of people did. Other such cults were much much worse because they turned rancid and violent- Synanon, Scientology, the Rajneesh crew in Oregon.
Yeah I've only heard about him in reference to Mother of All Demos. I've only heard that EST things very very recently, maybe because this review (while still in Google docs?) triggers an osmosis that led to Reddit?
Ok, now I have a better understanding: cancel culture was not a thing at the time. We did not judge people by cooties. There was probably no newspaper or magazine article about Engelbart that mentioned EST.
Thanks for linking to Gwern's essay.
While the contest review made me interested in this Project Xanadu thing, it also left me a bit confused about how exactly Xanadu (or memex) would be different from WWW. It was Gwern's essay which cleared up the confusion and also acted as a real review (i.e. saying whether Xanadu was good or bad).
I'm going to digress here and saying that this comment makes me actually see what gwern is, when before I just encounter their random articles linked here from time to time. And I can only say that, how could I not discover them much earlier than this? Even on this topic about Xanadu, I find their article much more competent and have personal touch. Especially because they actually seem to have tried to make something similar, resulting in their website having the actual implementation of transclusion that I actually like and hope is implemented everywhere else. ACX should go off Substack and hire whoever is making gwern's front-end.
The design article makes me explore their archives and discover that they write about things that are actually much much closer to my own interest, especially anime and Haskell. Rationality and psychiatry is nice to read sometimes, but nothing beats discussing Umineko, GAN, and Cabal. Looks like I found myself but better in everyway.
Even though i may not agree will all that they write and their writing is still not as excellent as SSC/ACX (well, it's a high bar), I think I may found a new blog to binge and it's all thanks to this comment.
I'd say that the biggest weakness of gwern.net is that it doesn't facilitate any kind of commenting/discussion. It's his own personal archive that he obscurely tinkers with. Sometimes it gets linked to, I enjoy reading whatever was linked, but with no way to engage it's not enticing to dive deeper.
Yeah it's an obvious flaw. I guess they'll have their own justification for this but I still miss it.
Small but kind of glaring error: the proximity fuse wasn't an anti-tank weapon in World War II, it was mostly used against air targets and secondarily against infantry. When you're trying to destroy a tank you want your shell to be intact when it impacts so it can penetrate the armor, not explode next to the target.
Another glaring error - "a one-line memo in June expressed Roosevelt’s total confidence in his Director: “Do you have the money?”
Indeed he did. The warheads it bought would fall on Hiroshima and Nagasaki in mere weeks."
If this was June 1945, then Roosevelt had been dead for two months. If it was an earlier year, then the bombs would not fall in mere weeks.
This is general for many of these reviews, but I feel a lot of them aren't even reviews, more like a summary of some interesting history. There's subjectivity at what, two points ("Xanadu looks cool" and "Nelson seems annoying")?
I guess the question is, does that matter?
many book reviews are reviews of histories or biographies, and are mostly summarizations of the stories within, sprinkled with some author commentary, who hopefully drags in other points of view, etc. this review seems like mostly a fairly standard book review of Nelson's autobiography, in this way, and i think if it were presented as such it wouldn't stand out as not being a "review"
Yeah next year might as well just call it a generic essay contest. Only maybe 2 of the 10 finalists so far have actually been reviews.
True, but this seems to be what people want. There were definitely actual reviews in the submissions and it doesn't seem like they were extra-likely to be chosen.
The review contest has always been an essay contest, much like the question time after presentation or a panel tends to be an "impromptu public speech about some preferably related topic" hour.
This is, of course, even moreso when you're not talking about something like a book that someone might, you know, want to read on the basis of the review (it's not like someone's going to be able to try out Xanadu-as-fully-imagined-by-Nelson on the basis of this review, after all)
I agree - the most highly regarded reviews appeal to the interests of the readers here and mimic the style of the substack reasonably closely. It makes sense: they come here to read text in a particular style on a fairly narrow range of topics.
Particularly with finalist reviews (of books) in the last year or two, none have motivated me even slightly to buy or read the book. (It could be argued that this isn't the goal of literary criticism - neither are the essays going to have much impact on authors and academics and future books, I expect.) So as reviews they fail, but then Scott is not in the business of selling books. As essays they succeed very well. They give an overview of a general area, normally with reference to what I assume is a reasonably good book, and maybe some insight into how thinking in the area falls into line with a rationalist world-view.
"The WWW had a number of advantages over Xanadu..."
I think this is important, but misses the most important advantage that the WWW had over Xanadu: Starting in 1993 people could actually use the WWW.
Followed quickly by: People were actually using the WWW.
Things that exist are more useful than vaporware.
"Real artists ship" - Steve Jobs
I have a little bit of a personal connection to Ted, as I was his neighbor in Sausalito for a bit. I moved onto a ferry boat called the Vallejo in 1994. You can see it clearly in this home video walkthrough by Ted and his son, near the end. https://archive.org/details/SouthFortyPier1988
He may have even come over for one of our infamous parties.
This is basically how I like to use llms to explore topics
Highlight text and either expand or navigate
"a baby-faced non-insane Elon Musk"
Not only is Elon not insane, but he is one of the most accomplished human beings on this planet, and in fact in history. Putting him alongside Edison, ... is entirely appropriate.
he seems pretty erratic lately with a some addictions and obsessions that seem to be destroying his health and wealth. whether he's a very accomplished person has nothing to do with it, unless you think that success immunizes you against mental problems
Addictions? lol - thanks for the diagnosis, doctor. After a full examination, of course.
if you don't want me to call these things addictions im happy to use whatever word you prefer but it wouldn't change the underlying fact that there are many reports that he is using ketamine frequently from people who know him personally. you can see how much he sleeps based on what times of the day he tweets: for years he was getting only a few hours of sleep per night.
Aren't you also making an assertion about his mental health, though...?
Yes, I'm assuming a man that runs Tesla, Spacex, Starlink, Twitter, The Boring Company for YEARS in excellent fashion and very profitably is in reasonable - at least - health.
I think it's inconsistent to say that somebody can't claim Musk has mental health issues based on available information about his behavior, but you can say he doesn't, based on available information about his behavior.
If redSun was a psychologist it would be unprofessional, but I think we can reasonably understand they're speaking colloquially.
No, "available information" re Musk's corporate/innovative excellence and personal wealth are widely available and agreed upon via the financial press.
Re drugs and mental health, there is contradictory rumor and slander based on politics, ...
Don't conflate the two. Entirely different.
His health, yeah, probably. But his wealth seems fine.
There’s definitely a lot of Edison aspects about him, between running and growing new technology industrial giants, and unnecessarily antagonizing many others and driving them away. (I don’t know if there’s anything that will be as much this anti-legacy of Musk as the Hollywood film industry was the result of Edison driving away all filmmaking from the New York area with his patent trolling.)
Ironic that Tesla is the century and a half old name that will end up being tied with Musk’s in future history.
I got a weird sense of deja vu reading this, I think dating back to a science fiction story probably 30 years old by this point, by an author that may well have been inspired by that WIRED article.
This bit especially: "Simply put a royalty on the links. If you want to reference a copyrighted New York Times article, then you’ve got to pay the author a little bit. And if someone else links to what you’ve written, then you get a small payout."
Except when I checked the series I was thinking of (https://en.wikipedia.org/wiki/My_Teacher_Is_an_Alien) it seems like they came out well before that article...so probably not. Either I'm thinking of a wholly different book or Bruce Coville is just a well connected guy, both seem possible.
Anyone else got a better lead?
After putting a few queries to ChatGPT (what would Nelson think of that I wonder), some other candidates are Earth by David Brin, or more in the YA line, Diane Duane: So You Want to Be a Wizard series, particularly High Wizardy since the manual comes in the form of a computer.
Probably burned a couple nickel's of compute to do the searching, although I doubt any portion of that would go back to the creators, alas Xanadu.
I used to read Bruce Coville's books as a kid, but never thought they had any connection to Project Xanadu.
Bruce Coville wrote a trilogy of science fiction novels, "The AI Gang," about AI development that he published in 1986 and then rewrote and revised in 1995. I haven't read them since the Nineties, but they sound more like what you were looking for than "My Teacher is an Alien."
I liked this. I’m happy I read it.
" There are no trails on the World Wide Web—instead, there are misattributed quotes, dead one-way links, constant plagiarism scandals, and widespread misinformation and mutual distrust. There are no trails on the World Wide Web—instead, there are misattributed quotes, dead one-way links, constant plagiarism scandals, and widespread misinformation and mutual distrust."
I'm not clear on how these aren't just two different ways of saying the same. The optimistic way, in which we have "trails" and a vague nebulous sense that they'll all be going something wonderful, worthy of a stately pleasure dome etc. And the pessimistic way. in which we have "links", many of which are dead and many of the live ones pointing to misinformation, distrust, plagiarism, and misattributed quotes. Why are the "trails" in Xanadu immune to the same forces that create all the crud in the current internet? If anything, from some of the description I'd expect them to be more vulnerable.
I'm getting the impression that "Xanadu" has only two significant advantages over the WWW. First, that it never managed to be implemented in messy reality, but could live only as dreams and promises, and second, that those dreams were frozen in place before Eternal September.
Indeed, the problem is people... and the complete lack of proposed technical defenses against the bad actors.
Writing an incorrect information is just as easy as writing a correct one, actually even easier because you don't have to waste your time fact-checking. The incorrect information could be corrected, but the correct information could just as easily be uncorrected... leading to edit wars? Ultimately the problem is that the words on the pages are unconnected to the facts on the ground, and anyone can assert any connection.
In a cooperative environment, "many eyes make all bugs shallow". In an adversarial environment, whatever method you use to correct the facts, the other side can mimic it to uncorrect them. For example, do you point out that a crackpot's page disagrees with hundreds of established sources? The crackpot will write hundreds of articles, and now your page disagrees with them. Soon there will be thousands of pages on homeopathy and quantum healing.
Then there are more subtle problems, such as adding a reference to a serious page, but misinterpreting its conclusions. Who is going to verify all such links? Also various ways of lying while being technically correct, that many journalists love so much. Cherry-picking. Etc.
It's thanks to the web being disconnected that we can have islands of relative sanity, such as ACX. Scott can set the rules about what is okay and what is not, without needing to get the approval of the rest of humanity.
Which reminds me: on web, you have not only static pages (with or without hyperlinks) but also web *applications*, such as discussions. And anyone can make their own. No mention of anything like this in Xanadu.
An insane story, and not in the oh-that's-cute way of figures like Emperor Norton. I guess it's good to know hallucinations were a problem with compsci long before LLMs? You can call it preconceived notions from growing up with the web (just barely old enough to have a before-and-after experience, one of first households in our town to get internet), but I think ending up with The Internet But For Schizophrenics would have been Bad, Acktually*. Are we really sure normal humans think in this sort of nonsystematic, freewheeling, trails-of-associations way? Are we really sure the blame for impersonal coldness and cruelty lies with the infrastructure that connects us, rather than PEBKAC? Never bet against PEBKAC! I just don't buy it. More limited claims, like social media specifically being poorly aligned, seem self-evidently true by now. I think you'd need to bring considerably more receipts to justify throwing out the baby with the bathwater though. Seems like one common sign of being Terminally Online is reaching the conclusion that "No, it's the *internet* that's wrong", yet not being able to log off it anyway.
Also, I regret following that youtube link for Suggested Reading Music. It made me distrust the moral aesthetic concepts (still think this is a weird Chomskyist conflation) behind the review even more, and this wasn't a particularly easy review to read through in the first place. Reversion to mean with previous years' contests: wild swings in review quality. I guess some things never change...
*although the pay-per-click/microroyalty thing would go a long way towards deterring spam and low-effort posting of all kinds, as well as elegantly solving the LLM plagiarism debacle, if it could be implemented
Pay-per-click would also deter e.g. people volunteering for Wikipedia. Imagine that each time you fix a typo, you also have to pay for the privilege!
The details lie in the devil, for sure...you'd want the price low enough (zero?) not to incur Trivial Inconvenience penalties for prosocial things like building wikis, yet high enough to make the David Gerards of the world think twice. I'm fairly convinced that very low charges like a mere penny would disrupt the business model for many kinds of automated spam...more advanced implementations would also include a "refund the charge for mutually agreed upon positive interactions". But that's already starting to sound pretty complex for just a single site to manage, nevermind do so internet-wide by default.
(Of course, in an alternate world of "internet is powered mostly by subscriptions rather than ads", perhaps such payment infrastructure would already exist. But we didn't end up with that version of the web.)
I have heard that at some moment there was some kind of "web on mobile phones" in Japan (not html, but some special protocol). Users paid per minute spent online, because that was considered normal with phones. And about 10% of the money you paid went to the author of the web page you were browsing, which was easily measured by the phone companies. As a consequence of this model, there were many web pages in this protocol providing various services. I think I was told that at some moment, the number of web pages in this protocol (which didn't exist outside of Japan) exceeded the number of web pages on the internet.
Unfortunately, it seems to me that a model like this wouldn't work well today, because companies like Facebook and Xitter would optimize even harder to make the users never leave their web pages. The problem is, this model incentivizes you to create content that people want, but also disincentivizes you from linking someone else's content. When you are a small player, linking to others still kinda adds value to your pages (people are more likely to bookmark a page "links to 100 useful resources on some topic" rather than those pages), but when you are a big player, such as a newspaper or a social network, you can provide alternative links to your own content instead.
I already find this annoying when I read online newspapers that typically when they talk about some online thing (let's say, Khan Academy), the only thing missing in the article is the fucking link to the thing they are talking about. Instead, there are dozen links to other articles from the same newspaper on similar topics, or even articles from the same newspaper on other topics. Ironically, sometimes it is difficult to find a web service that many newspapers talk about, even by Google search, because those newspapers often have greater page rank.
Fascinating! Yes, it'd definitely be doable at the "hardware level"...not duplicable today because you'd end up with a bifurcated internet where mobile viewing effectively costs more, and good luck changing that equilibrium. People hate giving up "free". (Although maybe that's Working As Intended?)
Solving the linkage problem is hard, and that's one part of the review I'd have wanted to see fleshed out more...what exactly stops the sort of "Google summary sidebar" treatment (even before the AI Overviews), where you show relevant info snippets all within your own sandbox, and as a result that eliminates most traffic to the actual thing? Sure, you could charge for that snippet - but it'd be very weird to charge as much for that as actually visiting the page in toto. We see the same problem with Twitter going out of its way to throttle external links, especially to Substack. A link doesn't do any good if no one actually follows it!
'Are we really sure normal humans think in this sort of nonsystematic, freewheeling, trails-of-associations way?' -- I do think in this way. It's the kind of stream of consciousness you experience while out walking without headphones or anyone to talk to; or while doing random tasks around the house. I feel it used to be almost a default state of mind before mobile phones meant that I constantly have some source of entertainment by my side. It concerns me that when I look around now e.g. on public transport that not a single person is daydreaming, or (if I'm more critical) thinking.
I'm sympathetic to the idea that the Web as designed is techno centric rather than human centred - it's about facts and figures, not human experience. Certainly it could have been developed in a different way. On an entirely different note: the web was developed for good-faith communication between professionals - this is the fundamental reason internet security is so hard. If someone would design a system from scratch today, it would be totally different. Inertia is the reason cyberattacks are so common.
I guess this is a Typical Mind thing...I have a separate subprocess that flits around associatively, usually choosing to play whatever music suits the mood, but is also responsible for things like reflexive pun generation. For Real Thinking(tm) though, that part gets sidelined, and we bring up Systematic Filing And Calculation. One could certainly say that's been influenced by the web - even though I don't do much with phones, obviously I'm online, and have never been too keen on nature retreat type stuff. (Not because of lack of internet, but because I just like lightbulbs, heat, running water, lack of bugs, etc...) So it's hard to do a true controlled experiment, see how my thinking would change if durably distanced from our cybernetic minds.
It is indeed depressing that most people in public seem hopelessly addicted to their phones. Not just on public transit, but walking the streets, or driving, or in the middle of crosswalks (and then they have the audacity to get mad when *they* bump into *you*), or standing by awkwardly and helplessly while you bag their groceries. Such a lack of being present in Simulacra Level 0, of potential human connections never made. I don't really wanna judge it as "not thinking", but...a lot of it certainly is passive consumption. However, the fact that it's apparently so addictive is part of what makes me question the Xanadu narrative - somehow this supposedly deeply inhuman internet manages to ensnare billions of people anyway. If it were really so alienating, many more of those people would be daydreaming, noticing the world around them (even doing this while listening to music is an improvement), seeing those of us "open people" on the bus wondering where all the thinkers went. Instead they while away their leisure watching Chinese spyware propaganda and stoking division for their fellow man...
This was a really interesting review and I can't help thinking I'd enjoy a biopic. A family member has worked in digital records management and some of the principles seem to cross over. If a public body is the potential target of a Freedom of Information request, it's in their interest to make files as accessible as possible and as quickly as possible - which means good naming conventions, version control, and metadata. Cue moaning from staff who want to have a file called Misc or Random.
You can pry Misc_Random_Final_v2.1_Real_Final_Dont_Delete_Copy(1).docx from my Cold Dead Hands.
Xanadu projects were still ongoing into the 2010s. Here's a neat article on some of them by one of the engineers who was still working in this space: https://hackernoon.com/an-engineers-guide-to-the-docuverse-d080cdbb73a6
A mention of the person who actually dreamed of the internet, etc. would have been useful:
"The Dream Machine," by M. Mitchell Waldrop. Subtitle "J.C.R. Licklider and the Revolution That Made Computing Personal."
Indeed!
The Internet Society has its own narrative, making Licklider a prominent figure.
https://www.internetsociety.org/internet/history-internet/brief-history-internet/
In a round-about way, modern systems with cross-links, especially shared notion pages and vast obsidian pages with peoples' dozen hyperlink plugins, *are* a lot closer to Xanadu than the normal web.
Yeah, Xanadu feels more like a prototype for a personal (or maybe organizational) knowledge management system, rather than for web in general.
The vision of this project sounds to me like a collabrative Zettelkasten, which is an idea that pops into my mind every so often.
Perhaps it won't ever work on the public web, but I do think that various groups and organizations can benefit from having their own intra-memeplex, in a similar way to the commonly found intra-nets.
Notecards at Xerox PARC - https://en.wikipedia.org/wiki/NoteCards - worked. Circa 1984. Perhaps should be part of the story.
I recall that Phillip Hallam-Baker once proposed "The reason the Web works and previous attempts to implement [Ted Nelson's Project] Xanadu failed is that the Web has 404 Not Found. Ted Nelson insisted on referential integrity and since achieving that is a byzantine generals problem, the systems never worked." Certainly the Web has a great advantage in that any cretin can do a marginally-adequate deployment of a web server.
Intriguing article! I think I see Vanevar Bush's vision, but I'm not sure I can see Xanadu's vision.
In one part you say:
"The three quickly figured out a new system that would allow users to reference and link to specific parts of a file—they called these links tumblers, and made them work with transfinite numbers."
What does this mean? I'm very familiar with transfinite numbers --- but I don't understand how it could relate to linking to specific parts of a file, at all. Anyone have any idea?
Thank you for this. Having worked in the Web for 30 years, largely inspired by one talk of Ted's I saw, I worry about loosing the origin story.
On my first read through your comment, I thought you were referring to a TED Talk, and was momentarily thrown for a loop. I may start referring to them as "Talks of TED."
If you're interested in modern ideas along the lines of the memex and Engelbart's "Mother of All Demos", I recommend Ink & Switch's essays: https://www.inkandswitch.com/
(The essays may be interesting to a general audience, but they are mostly aimed at software engineers, since Ink & Switch is very interested in how to actually build these things - including many prototypes and occasional spin-off products.)
Nigel Calder once posted this about a techno-prophecy by Maurice Wilkes in 1964 of a world information system on networked computers. Wilkes made no suggestions about how it might be implemented, certainly no mouse or hyperlinks, but Engelbart was not the only one keeping Vannevar Bush's vision alive.
https://calderup.wordpress.com/2010/05/04/internet-64/#more-150
And then we tried again and called it Arbital. Admittedly, Arbital worked pretty well, as-implemented, for an _extremely_ narrow use-case, which puts it well ahead of Xanadu.
Weirdly, this reminds me somewhat of TV Tropes, and how pages are linked there. Over at TVT, there is a large but finite list of tropes one can use to list the ideas that two works have in common, which sounds like what Xanadu/Mimex was trying to accomplish. Two pages on two different works list the tropes that are associated with each work, and viola, you have a set of ideas connecting those two works. Click on one of the listed tropes, and you get a list of other works that utilize that trope. TV Tropes is, therefore, a hyperlinked semantic network.
Not sure how that would apply to Xanadu (or the web), if it could be. But here's another potential weakness: TVT is notorious for being a rabbit hole. Click on any page, and several hours later you suddenly wake up, head full of the works and tropes you've been reading. Xanadu would be worse.
That’s an interesting point! I had long noticed that TVTropes was the one website that more reliably produces multihour binges than Wikipedia, but hadn’t quite understood why. If it’s something about the siren song of the infinitely recursive cross-linking that works both ways, that would make sense.
Yeah, I think it is. It helps you discover new works that have something in common with a work you already know, around an idea (trope) that you didn't even know was a thing. That sounds to me pretty close to the vision they had for this Xanadu thing.
The hook of the Memex is strong - that's a machine anyone who works with information would dream of having.
I think the review is built around how sad we should be that we got the WWW instead of the Memex ... but this seems like a false model.
The Memex is described as a personal knowledge management system - and in the last 20 years software tools (from Org Mode to Obsidian) have become pretty common.
I'd argue that Obsidian more or less fulfills all of the qualities that Bush described - maybe with needing to take the extra step of explicitly creating files to reference for each individual piece of work - article, book, video - that you consume. (I would argue this constraint is implicit in the description of the Memex since Bush describes the machine as physically containing copies of everything you've read).
So I'm not that sad Xanadu never came to book - or that we got the WWW instead.
My model
- The dream of Xanadu or the Memex makes a lot of sense for private information management - where bidirectional links and margin notes are super valuable. We have this today with Obsidian.
- The WWW makes a lot more sense for no trust global communications / content distribution networks. Xanadu would not avoid the problems with content on the WWW this review describes.
As someone who has certainly heard of Xanadu before (I remember first from a DEF CON video in particular) and got quite curious at least from the historical side, the author here has fine taste in my opinion. it should be noted that Ted Nelson actually had a YouTube channel for anyone interested, seemingly a lot of content in relation to this: https://www.youtube.com/@TheTedNelson/videos
From there in particular, there was a series of "Xanadu Basics" videos that explain quite in detail his actual thought processes for the system and how it generally functions, that's generally one of the only organized attempts I know of from them: https://www.youtube.com/playlist?list=PL0pAM29gc1g9o58LWnoXkqAVWq8y335hX
There was *also* on there a 1 hour and 40 minute long reply to "The Curse of Xanadu" article that was mentioned on this post: https://www.youtube.com/watch?v=ASgjSxNdDqI
Whether or not that's worth anyone's time here probably varies, I once again simply see this as quite interesting from a historical standpoint and nonetheless also think that, despite to what extent his ideas may have been considered redundant nowadays, the sheer quantity of his time at least with the topic of organizing information seemed promising to me. Quite nice to see this post pop up notification wise as a result.
I'm one of the many nerds who bought Nelson's books in the 80s and pored over them. I got to see Engelbart demo his system live just before Stanford turned off the one computer it ran on, mid 1980s. That's a treasured memory.
The math that makes neural networks possible was published in 1980. It wasn't until the 2010s that the gear was fast enough to run them at scale. People forget how mind-bendingly slow computers were before the mid-90s. Neural networks are what make associative computing feasible.
But! The one Big Idea missing from the early visionary days was Digital Money.
There's a lovely low-res game called "Kentucky Route Zero", an immersive journey through the Underworld of Appalachia. One of the sequences is using a buried computer mainframe which runs game-within-the-game: an unholy mix of Xanadu and the old Adventure game. KRZ is a great game to watch playthroughs of, you don't have to actually play it.
Computers were both slow and expensive, so even the stuff that Engelbart was doing (which did not require neural networks) wasn’t very practical from a cost perspective at the time.
Can someone explain to me what the difference would be between Xanadu and Wikipedia? I imagine decentralization across servers is one, but I'm looking for differences regarding the philosophy of hyperlinks, version control, and so on.
Copyright seems to be the biggest sticking point. Wikipedia is powered by the free/open sentiment, and with anything like a pay-per-use model it would've been a non-starter.
This is really interesting and fills in a lot of gaps of my knowledge. I studied under Andy van Dam and he often tossed around a lot of these names—nice to have the blanks filled in about how they all participated in the history of the Internet.
Tbh, I was surprised this article didn't address Urbit (https://urbit.org/) —since that seems like the modern attempt to completely rewrite the Internet into something better.
I really wish this went into the rise of Roam, Obsidian, and similar, because you can kind of see an in-between of Project Xanadu and the World Wide Web, especially in the use of back links
I'm a bit grossed out how people *want* the founders of the web they use to be Proper and British and Adult and Correct (and Male).
Idk about you, but I want a web founded by adhd ppl, by crazy kinksters, by trans folks, by nerds and enthuisasts and ... you know? people who have something to hide. They'll keep things save.
And, at the same time, people who don't hide. Because they won't be compromised.
Most Correct People are incorrect people in hiding, and I don't wanna live in their world.
So, do you like Xitter? It's certainly much crazier and less Proper and Correct than pre-takeover, and yet I have a suspicion that it's not your cup of tea.
huh what?
that's like the opposite of what I want.
I want to live in a world where people don't need to self-censor.
As far as I can tell, these days twitter is a place where nazi hate speech is very much permitted.
Thus, it's a place where vulnerable people have to self-censor heavily, or they get publicly abused otherwise.
Or in other words, it's not a place "to be improper", it's a place where you can openly be a nazi, or act in a way such that nobody can prove you aren't a nazi. The latter is what "being proper" often means.
And this is why I hate the "respactable ideas have to come from respactable people" way of thinking: It usually forces people to act like nazis-in-hiding.
Which means no non-nazi can tell apart their supporters and their enemies. And this gives hateful ideas more power over everybody's lifes. I very much want to live in a world where this cycle is broken.
And I start with refusing the idea that every non-cis-straight-mono-vanilla-english-christian-conservative-wealthy-binary-man has to do their utmost to hide that they are different.
Tolerance is a social contract. I don't tolerate ideas that don't tolerate me existing.
It's funny how your side pretty much controls the entire internet and still thinks that its persecuted, underdog bias is a hell of a drug. You think that people who disagree with you don't have to self-censor in order to avoid being branded a nazi?
Scolds have stopped being exemplified by The Correct And Proper British White Male a few decades ago, they're now the sort of people who use expressions like "ideas that don't tolerate me existing", but somehow it doesn't seem like we've come any closer to a world where people don't need to self-censor.
If Xanadu is so awesome, I expect it would see use in a few private companies, like a juiced-up Confluence. But I expect most private companies will find that the costs of maintaining a private docuverse just for them is too high for just a moderate benefit.
Is Xanadu one of those technologies that absolutely needs to have wide adoption and network effects before philistines like me can see its value?
The concluding sections of the essay are comparing existing software to vaporware. It’s not hard to make the latter sound better than the former.
Selection by association rather than indexing sounds profound, but it’s there in HTML, which allows links everywhere and doesn’t force a top down directory structure. Bidirectional links? Not built into HTML but Google implemented a “find all pages linking to this web page” search. The “lame text format and a lot of connected directories” criticism applies more to Gopher (which predated the World Wide Web and was outcompeted by it).
“See every quotation in its original context” requires some sort of enforcement mechanism to prevent people from quoting without providing context. “Quote without limit without permission” with an automatic royalty payment requires changes to copyright law, which would require renegotiating (or abrogating) international agreements. In an alternative timeline where Nelson actually managed to get a version of Xanadu built and released, he probably would have done it by placing these goals in a “to be addressed in future releases” bucket, where they might still be today.
I loved, loved, loved reading this. I have both the Mother of All Demos and the student film saved for later, and am considering getting the autobiography. I can only hope these things are half as interesting as the reviewer's descriptions of them.
Someone once wrote "The Web is wonderful and horrible -- it's like the really messy desk of a brilliant but deranged professor -- but who has a dim-witted graduate student who has more-or-less indexed everything. If you rummage around, you can find nearly anything."
What we *want* is Xanadu, that is, the Web, but everything properly organized, indexed, linked together, annotated about reliability, etc., etc.
What we *need* is a *smart* research assistant to organize everything on the Web. And it seems to me that it might be *just* within the foreseeable capabilities of an LLM-driven automaton to be that research assistant. For instance, if you want to judge the reliability of something, you need to compare it with similar things with known reliability, check how it's reviewed by sources with known reliability, etc., etc. This is what people do, but it seems within possibility that it could be automated.
Or perhaps, the Xanadu database itself isn't materialized but whenever you look at something, the automaton updates all the associated links. "Just-in-time Xanadu annotations".
That's a smart idea; apply some associative Xanadu style system to what already exists. The LLM analogy is suited; they seem to act much more like brains.
I really enjoyed this one.
I thought this was overall an excellent overview of some important history, if a little nastier than it needed to be.
One correction: You say "*Computer Lib* was his lamentation over the industry’s disdain for hypertext, and *Dream Machines* was Xanadu’s manifesto." I don't think that's an accurate summary, of either section.
*Computer Lib (You can and must understand computers NOW)* was intended as an overview of computer technology for non-technologists, including sections on programming languages, data structures, machine architecture, early hacker groups, and much more. The format and attitude was similar to the Whole Earth Catalog: look at all this cool stuff. Some strong opinions and even some lamentations, but that was not the dominant message.
*Dream Machines (New Freedoms Through Computer Screens)* was about the research frontier of interfaces and visions for human/computer interaction, of which Xanadu was just one example, It included eg very early CGI, real-time image processing, and ELIZA. It was not that different in form from the other section, just more forward looking.
I am not very objective about this book which had an enormous influence on me. I encountered it when it first came out, in 1975 or so. This was before the first personal computers were available and before there was any real mainstream idea that normal people should be interacting with computers at all. Both sections conveyed an irresistible air of excitement and revolution, and the idea that technology might actually be wrested from its military/corporate origins and put to humanistic uses. So, yes Ted Nelson may have some character issues and be not very good as a system builder, but he deserves enormous credit for his ideas and writing.
Thanks for this excellent reminder of computer history and how the a simpler approach like the Web can outperform a more kitchen sink approach like Xanadu. But you have to love Ted Nelson for how he really tried to get the average person into computers in an era where big iron dominated computing and computers were intimidating. Everyone needs to glance at Computer Lib and get a sense of how computing was in the 1970s and be really really glad to be living in a time of computers everwhere!
If you want to try note taking/word processing like this describes, try obsidian editor. It has a learning curve, but it's super powerful, and really nice to work with, and you don't need to know how to do everything to be able to use it well. It lets you super efficiently link to files (you can also link to sections of files) or create a new file as you type in a current one. I'm tired and not selling it well but if you read the idea of the memex and thought it would be cool to try, try obsidian or look up videos of people using it
The comments on ddos attacks are well considered and we did think about all that, and when I have a better solution for it I'll implement the system with it, for now I'm considering about reviving it. Roger Gregory
Your site is incredibly informative—thank you for sharing such valuable content. https://www.myfurnaceandcarpetcleanersedm.com/