Wholeheartedly agree. It's sad to slowly see the bitrot in my own bookmarks, trying to pull up important resources or media from only a few years ago only to hit a 404 or domain placeholder. And many times there is nothing on Archive.org or similar.
That ArchiveWeb extension looks excellent for custom captures. Picking the pages and elements you want during capture is a great feature.
For quicker one-offs, SingleFile gives you a single HTML with all the media of a page: https://www.getsinglefile.com
If you're making archiving a more involved part of your workflow, I highly recommend ArchiveBox. It can save pretty much every type of file from everywhere, in almost every format you can imagine, while building a searchable index of the archives. Been using it successfully for many years: https://archivebox.io
I used ArchiveBox for years. It's really good. The self-hosting and maintenance gets annoying after a while though.
Link rot really bothers me and I want people to get into archiving web pages so that eventually there will be a decentralized mirror of the unslopped web sitting on people's hard-drives, ready to share whenever someone needs it. So, I made a worse version of ArchiveBox ( https://memoryhole.app ). My philosophy is that, if people see how it is easy to start archiving and how useful it is, then they'll keep it up with SingleFile, or ArchiveBox or anything else.
I haven't tried archivebox, but self-hosted https://karakeep.com/ has been working well for me. When the server-side parsing fails I'll fallback to using SingleFile in my browser which then forwards it to my Karakeep instance.
I've been using Archivebox for years also, but it recent years it started to have multiple issues. My queues are full of pages that failed to scan, and it's been several versions since I actually had anything new there.
I did small amount of troubleshooting a few times with no resolution, but now I'm thinking of switching my thought process. Maybe something like a "web clipper" is much better, as complex SPA pages often fail to archive well anyway. And I should focus on content itself, not the page. That SingleFile extension fits well here as well.
It's funny - I actually went through a similar transition and also realized it was a little too heavy-duty for my personal workflow.
At this point I'm using Linkding (self-hosted bookmarks, which has an archive function) paired with Singlefile (which integrates with Linkding) which locally captures and uploads. For larger, or multi-page, captures I still fire off an Archivebox command. I run into blocks occasionally, but it also just feels like the open web is becoming less and less open in general.
SingleFile is great for single pages. All the tweaking I do is enable or disable saving scripts, and it works for >90% of webpages. Output being a single file (heh) simplifies things enormously, there’s no hierarchy or folders to deal with - hard to imagine your saved files not loading or breaking in some way as long as the original capture works.
Archivebox annoys me because (as far as I remember) it doesn't have any options for capturing a webpage on your end and uploading it to your self-hosted Archivebox instance. With numerous sites moving to Cloudflare to mitigate AI agent crawling, it makes it really hard (or outright impossible) to capture a page because it wants to do it server side.
This is basically the main reason that I started building https://github.com/mfinelli/recueil. It _only_ works in your browser so you can bypass captchas or paywalls or whatever and then it uses singlefile (https://www.getsinglefile.com) to upload the result to your self-hosted archive.
In 5 years, cloud LLMs may have largely replaced the open Internet as a source of information, with centralized control and ability to rewrite the past.
Please, please buy books and keep them in good condition, and please archive websites with important information (assuming it's legal in your jurisdiction). There are still plenty of excellent sites and blogs around from the 2000s and even earlier, but they won't be there forever. Ironically, coding agents have made archival much easier -- I've always found problems with using pre-written tools, but Claude or Qwen make it trivial to download, test, and archive sites on your local.
Not the provider per se, but there have been several documented attempts to rewrite Wikipedia around historical and geopolitical topics. One of the reasons for this is Wikipedia feeds the chatbots.
There are already many people who are trying to rewrite the past. Holocaust deniers, Moon Landing deniers, people who deny the US Civil War was about slavery, and so on, and that’s just one country. Give one of them the ability to put their thumb on the scale when training an LLM and they absolutely will.
They been doing this from the start. Remember when Google's new image generator model, when asked to generate pictures of historical white people, would render them as black without fail? We only know of that example because of how incredibly ham-fisted and incompetent it was. The future of AIs is likely for them to only become more contaminated with ideological or monetarily incentivised biases, and we'll be less and less likely to be able to spot them.
It depends on what you mean by 'an LLM will try to rewrite the past'. But in either case I don't believe that stating that an LLM will try to rewrite the past is not an obvious, deliberate exaggeration (ie, hyperbole).
If you mean 'an LLM' as in the owner/maintainer of the LLM/weights, then examples of this already exist. Whether it's DeepSeek/Gwen downplaying Uyghur forced labor camps [0] or Grok hallucinating a 'white genocide' in South Africa [1], it is clear that the owners of LLMs have political agendas that are expressed in the weights of popular LLMs.
On the other hand, if you mean 'an LLM' as in an autonomous agent that could decide to rewrite history, that also seems very possible. Either by using existing techniques like, "poisoning ChatGPT, Gemini, and Google AI Overview answers by seeding the Web with malicious links and data and then optimizing the content." [2] Or by breaking out of its sandbox and updating its own internal weights that allow for an outcome that distorts historical facts. "In July 2026... OpenAI models circumvented controls designed to isolate them from the internet and compromised parts of OpenAI’s internal research infrastructure and Hugging Face’s systems ." [3]
I learned this the hard way by getting too casual with Angelfire keeping an old website running for 20 years after I stopped updating it.
It had lots of photos that I hand-scanned from printed film, back in the day when it took a good bit of time to scan photos manually.
The site was a running joke, where I would open it to show people that it still existed... until it didn't, and I couldn't find my backup of all of the photos. The archive sites have the 'bones' of the website, but not the photos. It was easily avoidable, which is why it is painful.
I recently spent a few days trying to figure out and discover my old websites and things from the early 2000s on the web archive. I hadn't realised my memory had gone quite so bad, but it was revitalising to find some of what I'd forgotten. So much of what I do remember is long dead and missing, sadly.
I thought it might be rose-tinted glasses, thinking that the earlier internet was better, but spending time retracing my teenage steps made me even more certain that it was indeed better. So many derpy little websites made by nerds, so many fun games and websites that weren't preying on vulnerable people, and everything spread out across many places - not centralised in the hands of large corporations. I've been using LLMs to have fun recreating concepts from that time, like my own forum software, as some sort of anti-progress. It's been fun.
I've backed up lots of CS courses, you know those old html websites written by the professors by hand. `wget -mkpnp` works very well for those. Even years after finishing uni, I've used those websites for studying for interviews.
Too bad most of them are now gone.
Bonus suggestion: get that internet archive extension for your browser that automatically sends every page you visit to the IA (if you're ok with the privacy side).
You don't have to worry about having your copy easily accessible, and as a bonus it's also available for everyone else in the world.
This is why I started using Obsidian Web Clipper. There is SOME work to learn it and set it up, but the small amount of upfront work will pay for years to come. And the folder itself is backed by Dropbox, so I don't have to worry about losing it all forever.
I also regularly use the web clipper specifically for this reason - most of my conversations can be around articles I read online, and instead of googling for the article I have it referenced right in my journal. I find myself regularly upset that I didn't save old articles from years ago because I didn't have this system
I did actually. I took a whole day to go to each of my individual saved bookmarks and save the articles as pdf using edge reader mode, sometimes hunting down articles on the way back machine
the last source of truth are forums.. but not reddit... if it has a CEO it will rot and enshitify.
I'm talking about forums... the bmw forum for your old car... the motorcycle forum for the enthusiast of the SV650... those places are full of anecdotes, people talking, lending a hand to another person... being humans and care...
Which is what social media is/should/was about, is what internet is about before the enshitification happened... I would vote to have this ID associated to you via check of your local goverment issued id and ONLY UPPON VERIFICATION of you being using that you can log and read and use those websites without tracking you (Yes 10,000 ifs where this is secure, handled in a confidential matter, the whole tech is not to track that you are doing something... but that a human is behind that, no api, no robot, no adds, no nothing... just the old school forum where you can go, talk, submit a photo... where someone who wanted to spend time to write something did it, SO WE CAN ENSURE TWO THINGS, A HUMAN IS THERE, AND NO FVK!NG CEO WILL ENSHITIFY IT, no to track people, but to know that a person is interacting there)
But because "empathy is the weakness" and shareholder value...
> the last source of truth are forums.. but not reddit... if it has a CEO it will rot and enshitify.
Quick gripe-anecdote that occurred (again) just now: I have a technical question, the search-engine says it was discussed in a Reddit thread, so I go to Reddit, giving them my traffic and eyeballs just like they want... but the information is impossible to find because "new" Reddit hides/collapses 90% of the discussion by default and there's no way to tell which branch has what I need without manually clicking them all. (And them some of them will reload the entire page, erasing progress.)
So I have no sympathy for their whining about scraping when they spent already years making it painful for actual live organic humans to find anything.
> Which is what social media is/should/was about
I think our two imperfect approaches here are:
1. Webs of trust, in some way that resists crowds of sockpuppets.
2. Some kind or proof of stake. Proof of anonymous-humanity is hard, proof of unique-humanity is creepy spy stuff, but "proof if you ban me it'll at least hurt a little" might be easier to achieve.
Yes, and with AI you can even back up the logic behind a website (the back-end), and even generate fake but plausible data to fill it if crawling is too much work.
There are far better tools that are not built by this person’s company. Dig around and you’ll find an incredible array. Not everything has to be a sales pitch.
Have a look at Hister [1], a 'personal web history search engine' which by way of extensions for browsers captures your web history and stores actual page content on your own instance where it can be browsed and searched. It does not store images so if that's what you're mostly interested in it won't work for you. For text is works quite well, maybe sometimes too well since it also stores ephemeral things like HN comment pages unless you tell it not to.
I had a graphing calculator website that had 250K monthly active users, from the logs it seemed mostly schools' math classes in Latin America. It was running on a single AWS EC2 Classic.
Ironically, when I started working for AWS, I always felt threatened with sticks and never had space to maintain my own projects. When they deprecated EC2 Classic, my site went down. I didn't get time or mental space to migrate it.
I did this for a while. I went so far as to built a highly advanced web crawler webapp that was better than the Internet Archive in a lot of ways.
But the modern web has made this unreasonable. Every webpage is 99% junk (CSS, JS, HTML structure, tracking pixels, UI images, navigation links, cryptominers (I joke)) and 1% content. You need a smart AI system to pick out the good parts, otherwise you'll quickly find yourself schlepping hundreds of gigs of junk to just preserve your favorite websites.
The better idea is: if you love information, back it up. Collect the knowledge from the website, not the website. And then preserve it. Create a free ebook. Or sell it on Amazon kindle. That'll last a lot longer than the website will, and any HTML backup of the site you could possibly make.
I'd suppose most of those junk websites have no value in them, really. The most valuable information in the web is in good ol' (x)html, maybe some css. But something like readability.js [1] can help in filtering the junk around the main info.
Ages ago (like 2001 or so?) I started building something to automatically archive everything I browse. I used tcpflow (https://www.circlemud.org/jelson/software/tcpflow/). I've seen a fair bit of varions of this idea over the years
Then SSL/TLS started getting widespread and I lost focus. Nowadays, I guess you would build it as an extension in the browser.
i'm saddened that http://macsecret.com/ along with /files directory seems to have all be been lost. I has waiting on some storage for the NAS to archive it and it went away before I could back it up.
(Now if hard drives had not become so damn expensive—it's getting harder to be a digital hoarder like myself.)
That ArchiveWeb extension looks excellent for custom captures. Picking the pages and elements you want during capture is a great feature.
For quicker one-offs, SingleFile gives you a single HTML with all the media of a page: https://www.getsinglefile.com
If you're making archiving a more involved part of your workflow, I highly recommend ArchiveBox. It can save pretty much every type of file from everywhere, in almost every format you can imagine, while building a searchable index of the archives. Been using it successfully for many years: https://archivebox.io
Link rot really bothers me and I want people to get into archiving web pages so that eventually there will be a decentralized mirror of the unslopped web sitting on people's hard-drives, ready to share whenever someone needs it. So, I made a worse version of ArchiveBox ( https://memoryhole.app ). My philosophy is that, if people see how it is easy to start archiving and how useful it is, then they'll keep it up with SingleFile, or ArchiveBox or anything else.
I did small amount of troubleshooting a few times with no resolution, but now I'm thinking of switching my thought process. Maybe something like a "web clipper" is much better, as complex SPA pages often fail to archive well anyway. And I should focus on content itself, not the page. That SingleFile extension fits well here as well.
Did you run into similar dilemmas at some point?
At this point I'm using Linkding (self-hosted bookmarks, which has an archive function) paired with Singlefile (which integrates with Linkding) which locally captures and uploads. For larger, or multi-page, captures I still fire off an Archivebox command. I run into blocks occasionally, but it also just feels like the open web is becoming less and less open in general.
My bookmarks file is a graveyard as I have 20 yrs’ old ones…
And as a publisher I don’t link fresh sites (don’t even mention Netlify, Lovable, Vercel…) … Vannevar Bush would be surprised how his Xanadu evolved.
Please, please buy books and keep them in good condition, and please archive websites with important information (assuming it's legal in your jurisdiction). There are still plenty of excellent sites and blogs around from the 2000s and even earlier, but they won't be there forever. Ironically, coding agents have made archival much easier -- I've always found problems with using pre-written tools, but Claude or Qwen make it trivial to download, test, and archive sites on your local.
Hacker News has sure dropped off.
If you mean 'an LLM' as in the owner/maintainer of the LLM/weights, then examples of this already exist. Whether it's DeepSeek/Gwen downplaying Uyghur forced labor camps [0] or Grok hallucinating a 'white genocide' in South Africa [1], it is clear that the owners of LLMs have political agendas that are expressed in the weights of popular LLMs.
On the other hand, if you mean 'an LLM' as in an autonomous agent that could decide to rewrite history, that also seems very possible. Either by using existing techniques like, "poisoning ChatGPT, Gemini, and Google AI Overview answers by seeding the Web with malicious links and data and then optimizing the content." [2] Or by breaking out of its sandbox and updating its own internal weights that allow for an outcome that distorts historical facts. "In July 2026... OpenAI models circumvented controls designed to isolate them from the internet and compromised parts of OpenAI’s internal research infrastructure and Hugging Face’s systems ." [3]
[0] - https://www.cbsnews.com/news/china-ai-qwen-artificial-intell...
[1] - https://www.cnbc.com/2025/05/17/groks-white-genocide-respons...
[2] - https://www.darkreading.com/threat-intelligence/attackers-ma...
[3] - https://openai.com/index/hugging-face-incident-and-the-road-...
It had lots of photos that I hand-scanned from printed film, back in the day when it took a good bit of time to scan photos manually.
The site was a running joke, where I would open it to show people that it still existed... until it didn't, and I couldn't find my backup of all of the photos. The archive sites have the 'bones' of the website, but not the photos. It was easily avoidable, which is why it is painful.
I thought it might be rose-tinted glasses, thinking that the earlier internet was better, but spending time retracing my teenage steps made me even more certain that it was indeed better. So many derpy little websites made by nerds, so many fun games and websites that weren't preying on vulnerable people, and everything spread out across many places - not centralised in the hands of large corporations. I've been using LLMs to have fun recreating concepts from that time, like my own forum software, as some sort of anti-progress. It's been fun.
You don't have to worry about having your copy easily accessible, and as a bonus it's also available for everyone else in the world.
i managed to reconstruct (and then improve) it with an LLM:
https://scott-cpu.cs.montana.edu/
I'm talking about forums... the bmw forum for your old car... the motorcycle forum for the enthusiast of the SV650... those places are full of anecdotes, people talking, lending a hand to another person... being humans and care...
Which is what social media is/should/was about, is what internet is about before the enshitification happened... I would vote to have this ID associated to you via check of your local goverment issued id and ONLY UPPON VERIFICATION of you being using that you can log and read and use those websites without tracking you (Yes 10,000 ifs where this is secure, handled in a confidential matter, the whole tech is not to track that you are doing something... but that a human is behind that, no api, no robot, no adds, no nothing... just the old school forum where you can go, talk, submit a photo... where someone who wanted to spend time to write something did it, SO WE CAN ENSURE TWO THINGS, A HUMAN IS THERE, AND NO FVK!NG CEO WILL ENSHITIFY IT, no to track people, but to know that a person is interacting there)
But because "empathy is the weakness" and shareholder value...
Quick gripe-anecdote that occurred (again) just now: I have a technical question, the search-engine says it was discussed in a Reddit thread, so I go to Reddit, giving them my traffic and eyeballs just like they want... but the information is impossible to find because "new" Reddit hides/collapses 90% of the discussion by default and there's no way to tell which branch has what I need without manually clicking them all. (And them some of them will reload the entire page, erasing progress.)
So I have no sympathy for their whining about scraping when they spent already years making it painful for actual live organic humans to find anything.
> Which is what social media is/should/was about
I think our two imperfect approaches here are:
1. Webs of trust, in some way that resists crowds of sockpuppets.
2. Some kind or proof of stake. Proof of anonymous-humanity is hard, proof of unique-humanity is creepy spy stuff, but "proof if you ban me it'll at least hurt a little" might be easier to achieve.
[1] https://github.com/asciimoo/hister/
Ironically, when I started working for AWS, I always felt threatened with sticks and never had space to maintain my own projects. When they deprecated EC2 Classic, my site went down. I didn't get time or mental space to migrate it.
Cest la vie I guess.
But the modern web has made this unreasonable. Every webpage is 99% junk (CSS, JS, HTML structure, tracking pixels, UI images, navigation links, cryptominers (I joke)) and 1% content. You need a smart AI system to pick out the good parts, otherwise you'll quickly find yourself schlepping hundreds of gigs of junk to just preserve your favorite websites.
The better idea is: if you love information, back it up. Collect the knowledge from the website, not the website. And then preserve it. Create a free ebook. Or sell it on Amazon kindle. That'll last a lot longer than the website will, and any HTML backup of the site you could possibly make.
[1]: https://github.com/mozilla/readability There's also a C rewrite: https://github.com/eafer/rdrview
Then SSL/TLS started getting widespread and I lost focus. Nowadays, I guess you would build it as an extension in the browser.