I vociferously suggest that you add an introductory paragraph to this post that describes the goal of this project in terms suitable for a newbie that do not include acronyms. I had to read about 500 words before I halfway understood what you were talking about.
It sounds like crowdsourced/Wiki-like fact checking. Am I right? I hope your dream comes true. Good luck.
One function that you totally left out was copy editing and proofreading. I have done this professionally, and you still might find it necessary. Of course, compensation would lead to increased enthusiasm and investment of time.
Love the spirit of this. How would the project address the fact that people read more on their phones and there's no equivalent to Chrome Extensions there?
I don't know how to code, and I don't have a large audience, but I'd be happy to help if I can. Put me on the list of potential jurors if you need to scrape up some.
How does this meaningfully distinguish itself from Ground News? They already utilize AI in their summaries (I think) and offer access to competing news articles, along with its political bias.
Good question. At scale it’s because we would have a UI layer that follows you around your browsing experience. Secondarily, at scale we would financially incentivize participation and keep trust scores for various individuals/groups. So basically record keeping, rules, money.
What’s your strategy for user acquisition for Phase 1? It will likely be tough to scale beyond that Phase unless you have users constantly testing, finding bugs, and perhaps paying for your product.
Basically, get media influencers who care about the news to buy-in. Give them the ability to hand write headlines, even if that UI is clunky at first, and have them to advertise it to their audiences, maybe give editing authority to a few folks within their audience. That’s a slow fly wheel until we can get to the point you can fight it out with someone and then have a group of uninvolved people pick the winner. That’s high conflict so high interest and I think that is the point at which it would really ramp up.
I’m not sure I understand the purpose of what you’re doing. As in, if we’re allowing people to hand write these replacement headlines, won’t that just institute a different form of bias? It’s not exactly clear from this and your other articles, but I was under the impression that it will be an AI rewriting the titles of articles.
If you’re already doing a browser extension, consider copying RocketReach’s strategy (or was it Zoominfo or Apollo?) to have paid users, and contribution based users. If you want to pay you get access to their service of having the contact information of people from LinkedIn or wherever. If you don’t want to pay, you have to install this extension that scrapes your email for email addresses and names to add to their database. It’s actually super effective since that builds the most accurate and largest base of contacts.
Consider contributions to the community notes of the internet the criteria for access to the unpaid version, and maybe there’s a clear path to monetization for charging people not to have to contribute to those community notes, so to speak.
Rooting for you and all those involved with the project! I doubt I'd be much help with what you're currently working on (web dev's something I'm working on learning, but I'm still a neonate there), but I'll give a standing offer to help if you ever need gruntwork. Might be able to help more with Phase 2 repository work too, when you get to that.
I'm not completely sure I understand what you're trying to do here. It sounds like a system for editing news articles like a wiki, with a network of wikis with their own groups?
Crowdsourced news has been seriously tried several times by groups determined to make it work, but they failed. Using existing news sites as a base makes starting easier (Google's Sidewiki and some similar projects were popular for a little while, iirc), but there's no real way to verify their content anyway. If the underlying info isn't publicly accessible, it doesn't matter if a hundred people look over the article if the person who gathered the info simply lied, as certain prominent media sources have recently developed a tendency of doing.
I admit I would probably trust a bunch of random people on the internet more than I trust the NYT, but probably not by enough that I would consider it a solution to the problem.
I know I probably seem weirdly confident about this and I’ll explain in some more detail. But first, a quick explainer of why I think this is different than the other systems but most closely aligned to Community Notes.
All the other systems I’ve seen do not have the potential for an attention flywheel because the UI is wrong, except for Community Notes. Meaning, you can create some sort of jaw-dropping bombshell report or stunning refutation of an areticle and it just… sits there. Take ground.news as a for instance. There’s lots of really great information there but it doesn’t “get in your face” so to speak. There was an old SNL skit where the writers combined the show “Unsolved Mysteries” with particle physics, and speculated that someone had created a unified theory of everything and then left it in a briefcase that nobody could find anymore. The joke being that someone did some kind of stunning incredible work and did a really bad job of promoting it. What I mean by that in this context, is when someone sees a bad headline or terrible information, it’s already too late for the refutation to matter to something like 90% of people. The propaganda has already been delivered to the critical mass. Unless you happen to go from one article to the other refuting the first one, spend time thinking about which is correct, the two are wholly separate and relying on someone being a news junkie to do individual sense-making. That does not scale to the group level. We should be taking the hard work people do in one of these systems and promote it up so that everyone can benefit.
Community Notes is the first thing that has ever gotten the basic premise and the UI right. If you want to correct something, then the correction needs to be directly on the thing you’re correcting. Impossible to see one without the other. It also has the right back-end process, where it’s not some company with shady financial connections vetting whether or not the note is helpful but politically unaligned people, randomly chosen. This is what I mean by creating anti-spin. You’re not taking one particular group’s agenda at the emergent level, you’re canceling that out across groups that out to arrive at something that’s closer to truth on average.
The UI I want to build looks a lot like FoxVox https://palisaderesearch.org/foxvox (I think you may know the people involved? I tried getting ahold of them but i might seem like a random nut job). When you look at a news site, it would come prefiltered for garbage information. Instead of the right/left lens here, it would come with the lens you think is most appropriate, plus a whole bunch of other stuff that happens from inter-group conflicts and reputations coring. None of these changes would be hidden the way they are with FoxVox. If we change words/headlines then the change itself is visible through coloring or something and you’d be able to click a link to see why the change was made. The most valuable of the things you’ll see is what I’m thinking of as “Golden Notes” meaning notes that have survived inter-group conflict. One group promoted something that was disputed and was able to make a case that they were correct and multiple groups signed off on it being right. That’s as close as I can figure to getting something like adversarial audit working in a real-time news environment. The outcomes of all of these adjudications are saved and go into trust scoring metrics for the news organization and for the journalist.
You start with a group because you have stuff you care about. You have your internal group noting that you all follow and look at. But when you want to push something “Across” groups then you have to prove it out. You’ll be forced to do that for some number of things and because of that group visibility will either rise or fall due to the reputation scoring, so that everyone is always in competition to make widely understandable explanations to everyone else.
I think this can be successful where the others are not because it’s dramatic. You go into someone else’s territory, the NYT article itself for instance. You write all over the page how it’s full of crap. It all gets adjudicated publicly, so if you get some famous YouTube taking one side and another taking the other, it gets eyeballs, but then the whole of the evidence is reviewed by people who weren’t hanging around to look at that one specific issue. There’s winners and losers. There’s human consequences. There’s lots of loud flashily arguments. Group honor is at play.
That’s all if this works. It very well might not. It might not ever hit that critical threshold of user engagement. But there’s a flywheel here, I think, where if you poured enough money into the system to get it started, get people to see the concept working, that it then takes off on its own. We trust criminal justice to juries. We use scoring systems to extend credit. Why not news?
I work with a pretty complex decision engine in my day to day which is why I feel pretty confident I could make this work, technologically. The ugly pieces with reputation, money movement, etc, when it’s at scale are all things I already do on a daily basis. That’s the magic I think would this sing and it’s the thing I have the most experience with. Getting the user base going and buying the right attention is a whole different can of worms and I’d be counting on luck to get it up and going. I already see people doing this basic format everywhere though. Tracingwoodgrains is facilitating some bets on X and it’s the same kind of thought process. Put your money where your mouth is, make an argument, let someone else decide. You get some of this with prediction markets but we need something different for news, although I’d love if polymarket started shoving that kind of contextual information in as well. Anyhow, I’m sleep deprived and have probably rambled. If you want to know anything more about this please let me know.
- For #1, rather than scraping known sites, could we have a UI element to request a headline for a previously unseen article? It shouldn't take that long to generate and cache a new headline.
- For #9, a very long-term out-there concern is the danger of creating filter bubbles. In Karl Schroeder's _Lady of Mazes_, the VR system blocks out people and things that aren't in your approved subset of reality, which I think is contrary to the ethos of this project. I would like to see more perspectives on things rather than one approved narrative, however much I agree with it.
On #1 are you thinking a “generate only demand” type of scenario? I can see that working but to show power of the system I would want to be able to demo whole site rewrites, for instance there are some clickhole sites where it would be more economical to just do the whole thing. And some may prefer to just have a tactful summary headline for everything.
In #9 I’m in full agreement. The way this should work is that there’s the original text, your groups view, and a consensus view that has “won” in some sense. Everyone is competing for that consensus view and that’s what causes the filter bubbles to pop so to speak and incentivizes the groups to be truthful and honest because theyre no longer trying to please only their own side. We should have a call and discuss more. Mornings are best for me if that is workable to you.
A request: before diving into this project and writing things from scratch, I’d like to see past work, existing work, and possible competitors. We are not the first to think of this idea. We need to know why previous attempts failed so we don’t try something similar and waste everyone’s time. This will save us tens of thousands of dollars in the medium and long run.
I know community notes is similar to this. Why hasn’t this scaled beyond twitter? Who has tried it before?
As Gandalf said: “Questions… questions that need answering!”
I vociferously suggest that you add an introductory paragraph to this post that describes the goal of this project in terms suitable for a newbie that do not include acronyms. I had to read about 500 words before I halfway understood what you were talking about.
It sounds like crowdsourced/Wiki-like fact checking. Am I right? I hope your dream comes true. Good luck.
One function that you totally left out was copy editing and proofreading. I have done this professionally, and you still might find it necessary. Of course, compensation would lead to increased enthusiasm and investment of time.
Good call on the intro. I assumed most regular readers had the background. I’ll see if I can’t add something.
on the copy editing and proofing front, hopefully as we expand that will become feasible.
Love the spirit of this. How would the project address the fact that people read more on their phones and there's no equivalent to Chrome Extensions there?
Long term, we’d build it for mobile compatibility. This is just proof of concept.
I don't know how to code, and I don't have a large audience, but I'd be happy to help if I can. Put me on the list of potential jurors if you need to scrape up some.
How does this meaningfully distinguish itself from Ground News? They already utilize AI in their summaries (I think) and offer access to competing news articles, along with its political bias.
Good question. At scale it’s because we would have a UI layer that follows you around your browsing experience. Secondarily, at scale we would financially incentivize participation and keep trust scores for various individuals/groups. So basically record keeping, rules, money.
What’s your strategy for user acquisition for Phase 1? It will likely be tough to scale beyond that Phase unless you have users constantly testing, finding bugs, and perhaps paying for your product.
Basically, get media influencers who care about the news to buy-in. Give them the ability to hand write headlines, even if that UI is clunky at first, and have them to advertise it to their audiences, maybe give editing authority to a few folks within their audience. That’s a slow fly wheel until we can get to the point you can fight it out with someone and then have a group of uninvolved people pick the winner. That’s high conflict so high interest and I think that is the point at which it would really ramp up.
I’m not sure I understand the purpose of what you’re doing. As in, if we’re allowing people to hand write these replacement headlines, won’t that just institute a different form of bias? It’s not exactly clear from this and your other articles, but I was under the impression that it will be an AI rewriting the titles of articles.
If you’re already doing a browser extension, consider copying RocketReach’s strategy (or was it Zoominfo or Apollo?) to have paid users, and contribution based users. If you want to pay you get access to their service of having the contact information of people from LinkedIn or wherever. If you don’t want to pay, you have to install this extension that scrapes your email for email addresses and names to add to their database. It’s actually super effective since that builds the most accurate and largest base of contacts.
Consider contributions to the community notes of the internet the criteria for access to the unpaid version, and maybe there’s a clear path to monetization for charging people not to have to contribute to those community notes, so to speak.
If you read this piece it will help see the full vision
https://extelligence.substack.com/p/how-to-make-an-information-super/comments
I will not let my lack of programming skill or audience get in the way of being involved with this project.
Shooting a dm your way!
Rooting for you and all those involved with the project! I doubt I'd be much help with what you're currently working on (web dev's something I'm working on learning, but I'm still a neonate there), but I'll give a standing offer to help if you ever need gruntwork. Might be able to help more with Phase 2 repository work too, when you get to that.
I’ll add your name to the list!
I'm not completely sure I understand what you're trying to do here. It sounds like a system for editing news articles like a wiki, with a network of wikis with their own groups?
Crowdsourced news has been seriously tried several times by groups determined to make it work, but they failed. Using existing news sites as a base makes starting easier (Google's Sidewiki and some similar projects were popular for a little while, iirc), but there's no real way to verify their content anyway. If the underlying info isn't publicly accessible, it doesn't matter if a hundred people look over the article if the person who gathered the info simply lied, as certain prominent media sources have recently developed a tendency of doing.
I admit I would probably trust a bunch of random people on the internet more than I trust the NYT, but probably not by enough that I would consider it a solution to the problem.
I know I probably seem weirdly confident about this and I’ll explain in some more detail. But first, a quick explainer of why I think this is different than the other systems but most closely aligned to Community Notes.
All the other systems I’ve seen do not have the potential for an attention flywheel because the UI is wrong, except for Community Notes. Meaning, you can create some sort of jaw-dropping bombshell report or stunning refutation of an areticle and it just… sits there. Take ground.news as a for instance. There’s lots of really great information there but it doesn’t “get in your face” so to speak. There was an old SNL skit where the writers combined the show “Unsolved Mysteries” with particle physics, and speculated that someone had created a unified theory of everything and then left it in a briefcase that nobody could find anymore. The joke being that someone did some kind of stunning incredible work and did a really bad job of promoting it. What I mean by that in this context, is when someone sees a bad headline or terrible information, it’s already too late for the refutation to matter to something like 90% of people. The propaganda has already been delivered to the critical mass. Unless you happen to go from one article to the other refuting the first one, spend time thinking about which is correct, the two are wholly separate and relying on someone being a news junkie to do individual sense-making. That does not scale to the group level. We should be taking the hard work people do in one of these systems and promote it up so that everyone can benefit.
Community Notes is the first thing that has ever gotten the basic premise and the UI right. If you want to correct something, then the correction needs to be directly on the thing you’re correcting. Impossible to see one without the other. It also has the right back-end process, where it’s not some company with shady financial connections vetting whether or not the note is helpful but politically unaligned people, randomly chosen. This is what I mean by creating anti-spin. You’re not taking one particular group’s agenda at the emergent level, you’re canceling that out across groups that out to arrive at something that’s closer to truth on average.
The UI I want to build looks a lot like FoxVox https://palisaderesearch.org/foxvox (I think you may know the people involved? I tried getting ahold of them but i might seem like a random nut job). When you look at a news site, it would come prefiltered for garbage information. Instead of the right/left lens here, it would come with the lens you think is most appropriate, plus a whole bunch of other stuff that happens from inter-group conflicts and reputations coring. None of these changes would be hidden the way they are with FoxVox. If we change words/headlines then the change itself is visible through coloring or something and you’d be able to click a link to see why the change was made. The most valuable of the things you’ll see is what I’m thinking of as “Golden Notes” meaning notes that have survived inter-group conflict. One group promoted something that was disputed and was able to make a case that they were correct and multiple groups signed off on it being right. That’s as close as I can figure to getting something like adversarial audit working in a real-time news environment. The outcomes of all of these adjudications are saved and go into trust scoring metrics for the news organization and for the journalist.
You start with a group because you have stuff you care about. You have your internal group noting that you all follow and look at. But when you want to push something “Across” groups then you have to prove it out. You’ll be forced to do that for some number of things and because of that group visibility will either rise or fall due to the reputation scoring, so that everyone is always in competition to make widely understandable explanations to everyone else.
I think this can be successful where the others are not because it’s dramatic. You go into someone else’s territory, the NYT article itself for instance. You write all over the page how it’s full of crap. It all gets adjudicated publicly, so if you get some famous YouTube taking one side and another taking the other, it gets eyeballs, but then the whole of the evidence is reviewed by people who weren’t hanging around to look at that one specific issue. There’s winners and losers. There’s human consequences. There’s lots of loud flashily arguments. Group honor is at play.
That’s all if this works. It very well might not. It might not ever hit that critical threshold of user engagement. But there’s a flywheel here, I think, where if you poured enough money into the system to get it started, get people to see the concept working, that it then takes off on its own. We trust criminal justice to juries. We use scoring systems to extend credit. Why not news?
I work with a pretty complex decision engine in my day to day which is why I feel pretty confident I could make this work, technologically. The ugly pieces with reputation, money movement, etc, when it’s at scale are all things I already do on a daily basis. That’s the magic I think would this sing and it’s the thing I have the most experience with. Getting the user base going and buying the right attention is a whole different can of worms and I’d be counting on luck to get it up and going. I already see people doing this basic format everywhere though. Tracingwoodgrains is facilitating some bets on X and it’s the same kind of thought process. Put your money where your mouth is, make an argument, let someone else decide. You get some of this with prediction markets but we need something different for news, although I’d love if polymarket started shoving that kind of contextual information in as well. Anyhow, I’m sleep deprived and have probably rambled. If you want to know anything more about this please let me know.
Thank you for the intro to Buckaroo Banzai
I loved that film.
Add me to the Discord!
I will message!
A couple of random thoughts:
- For #1, rather than scraping known sites, could we have a UI element to request a headline for a previously unseen article? It shouldn't take that long to generate and cache a new headline.
- For #9, a very long-term out-there concern is the danger of creating filter bubbles. In Karl Schroeder's _Lady of Mazes_, the VR system blocks out people and things that aren't in your approved subset of reality, which I think is contrary to the ethos of this project. I would like to see more perspectives on things rather than one approved narrative, however much I agree with it.
Thanks Gordon
On #1 are you thinking a “generate only demand” type of scenario? I can see that working but to show power of the system I would want to be able to demo whole site rewrites, for instance there are some clickhole sites where it would be more economical to just do the whole thing. And some may prefer to just have a tactful summary headline for everything.
In #9 I’m in full agreement. The way this should work is that there’s the original text, your groups view, and a consensus view that has “won” in some sense. Everyone is competing for that consensus view and that’s what causes the filter bubbles to pop so to speak and incentivizes the groups to be truthful and honest because theyre no longer trying to please only their own side. We should have a call and discuss more. Mornings are best for me if that is workable to you.
Your assigned tasks are beyond my scope & energy level, but I can offer my trusted news & truth sources...
Heck yeah
Some of this reminds me of what Epsilon Theory is doing with their Fiat News project: https://www.epsilontheory.com/fiat-dashboard/
And Phase 9 reminds me of They Live.
That movie gave me a lot of eureka moments around what the UI should look like.
The gears are turning!
A request: before diving into this project and writing things from scratch, I’d like to see past work, existing work, and possible competitors. We are not the first to think of this idea. We need to know why previous attempts failed so we don’t try something similar and waste everyone’s time. This will save us tens of thousands of dollars in the medium and long run.
I know community notes is similar to this. Why hasn’t this scaled beyond twitter? Who has tried it before?
As Gandalf said: “Questions… questions that need answering!”
Are you free most mornings? We should have a call. I have done background on some of this already.
God help us if you have another paternity leave....
It remains to be seen if we can get this to go anywhere, so I might stress myself out then for no reason as well.
I care about the truth and love to argue. And bake bread. So...
What kind?
I favor a crusty no-knead kind of bread, my bread guru is found at https://alexandracooks.com/
No bandwidth to help with development, but I'd love to be a tester.