I’m Putting a Team Together

Building the Trust Assembly

A very helpful commenter pointed out that it would have been best if I had included a link to this full description of the system I’m talking about. Basically, high level, it’s community notes for the whole internet.

Firstly, forgive any typos. I promise that in my sleep-deprived state I at least hallucinated that everything looked correct at the time it was written.

Quite a few folks reached out to talk about building a Trust Assembly over the last few weeks. And many of you are extremely talented! We already have a Discord going with several developers to discuss plans for building a MVP. As fate would have it, the uptick in attention also coincided with the birth of my second child. What I’ve written since that time, I have written with my thumbs on my cellphone during bath time, nap time, stroller time, burping time, etc. Nevertheless, we must go on if we are to have anything at all like a good future.

If SpaceX can land a booster the size of a 30 story building, I can figure out a way to write requirements for a phase zero project. My time is also becoming more free-up-able thanks to the surprisingly large number of you who have subscribed. My wife, who doesn’t really follow any of this stuff too closely, was very surprised and pleased by this.

Here’s the plan at present with phases, etc.

Phase 1: Headline Automation and Database Creation (We are Here)

Use an LLM and samples of writing from various creators to create an automated headline replacer. Scrape text from some predetermined sites, such as CNN, Fox, NYT, etc. Scope will be small so we could even start with just those three or even only one as a demonstration. Use the text to generate a new headline and store this in a back-end database. Build a chrome extension that takes the new headline and replaces it with appropriate UI changes to make the editing visible to the end-user.

Status: I think we have all the people we need to do the scraping, LLM work, and database creation. We could use some more experienced UI folks if anyone is interested. There is no funding as of yet for this project.

As soon as we have something to show, we can email a bunch of news folks and ask them if they are interested. Parallel to that effort, we can use those commitments to seek some funding for the project.

Phase 2: Editing and Forum Creation

Before we go live to end-users we need to build in a basic level of editing ability for headlines. Say the LLM gets it wrong, which will definitely happen. If you’re one of the initial system contributors now you have brand damage. So the initial contributors would need the ability to edit the headline. However, as stated previously, justification has to be given. So we need to build both the edit ability and repository/wiki-like forum for that justification to be housed as we build out discussion.

As this continues, we would expand the number of sites being scraped and automated but still on a fixed list to control costs and complexity. We would also in parallel expand the number of contributors with deliberate efforts made to create diverse polities within the network.

Phase 3: In-Line Edits and Annotations

Expand the editing abilities to include in-article edits and annotations with the appropriate UI displays. In parallel to this, give the contributors the ability to delegates his out to other users within their polity. I think this is the best way to start building out the user-base initially. It will evolve as the system evolves but we want heavily engaged epistemology nerds for our first users.

At this point, we can also bring in people who have specific issue focus that have a strong desire to refute specific pieces of information that mainstream media gets wrong or does a poor job on.

Phase 4: Group Sharing, Group Conflict, Group Adjudication

At this point, we would hopefully have enough people for this to be worthwhile. Plans never work out exactly the way you’d hope, but I like to have stuff written down. Give groups the ability to “inherit” the notes of other groups if they seem superior. It will all have to be gate-kept initially by the first contributors and their delegates but we want to start building that functionality out. So the group leader would have the ability to do something like say “Inherit all the notes of this group.” We also want groups to be able to find other notes at this point, a sort of “view all mode” for contributors, and then argue them publicly. This would expand our wiki-like database for this activity and help us build out functionality there. We could also get additional influencers who work in debate style formats interested at this point. We can let people argue it out on YouTube, via text, etc. But we might get more attention if there are high profile arguments on X or on Substack, etc. Enable the basic jury functionality to work asynchronously. Depending on how many people we have it might be the case we need to let the arguments happen and wait to get the response back.

Phase 5: Open Onboarding, Open the Adding Additional Sites

This is the point where we have enough “stuff” that we can meaningfully add people without having to vet them first. Let them create their own groups without having to go through us, etc. We would also want to build out flexibility so that someone could add any site to the automation list or automate any site not in our preset list. There might still need to be some limiters here and we would work it out depending on volumes, but balance here is that we want people to be able to argue for the truth but we don’t want to become Orwellian and rain on the parade of someone’s uncle who thinks Hillary Clinton is a warlock and has 12 followers, three of which are his own alts.

Phase 6: Reputation

Add reputation to all the groups, contributors, etc. There are so many ways to do this but basically, I want to just try different scoring methods and get large groups of people to look at them and vote on which one looks right. The whole point of the system is to create something people trust and that looks right to them. This is also where we are going to start posting policies, do outreach to professional associations, etc. For instance, ghost editing articles is just a terrible practice. That’s where someone publishes something, is called out for a mistake, then fixes it but doesn’t note the article that they made the edit later. That’s obviously wrong. We need a rule book somewhere around these practices.

At all times, I also want to make sure that we incentivize people to be honest about making mistakes and fixing them in a reasonable amount of time. I don’t want to drop a hammer on someone unless they’re really asking for it.

Phase 7: Money plus Reputation

Okay, now we need to get into some quite involved relationships with a bank. We’ll need someone who can set-up escrow accounts for us when we use the reputation established in phase 6 to start allowing people to pay for disputes. Then we need them to sit on the money, wait for the winner, and make a payment out to an account.

We want to push as much of the regulatory requirements here to the bank as possible. We take a fee upfront, prior to the escrow, and use that to pay ourselves, our jurors, and the people representing the sides. This is where you win if you’re right and you lose if you’re wrong, and all kinds of other stuff that will need a whole Legal team sitting around eating shotgun shells for breakfast to decide.

This is also our opportunity if we have funding to buy engagement and prove out the system. This is also why we need to make sure we get some real epistemology nerds in our userbase so that we set a good tone for further engagements. We’ll also need rules like a court for objections, evidence, etc. We’ll need to build up a culture for all of this in Phase 4 so that by the time we get here it is all established and understood.

Phase 8: Enterprise Integration, API, AI Data Feed

I think the best future for a product like this is for it to be offered as part of your normal newspaper subscription. We own the back-end tech pieces and we maintain a few browser extensions as a hedge against being shut out. We get the NYT, WaPO, and others to adopt this at scale, which probably involves some additional work to make something like UI’s for them but hopefully we get to nibble some of their subscription revenue so they can sell stuff like “Reputation Insurance” to their customers.

We’ll probably do this in earlier phases, but we will also hopefully get many customers like social media sites using our API to enhance the feeds of their users with our information. We’ll want to track all of this data so our writers know how much their work is spreading, etc.

We will also probably be able to attract a strong customer interest from AI companies both for training data and for just the enrichment in search we could offer. We will take a cut of this but the idea is that the money primarily goes to our contributors. The exact split is something that I again want to decide democratically but would push for the highest value work where we have a big delta between what was believed and what was proven but he system is rewarded most.

If we get enough users, it should be an easy sell to get large corporations to make standing accounts with us to pay for priority on their disputed items and to make sure their escrow accounts are immediately funded. We might even be able to do this earlier. A lot of it is a “It depends” kind of thing.

This is hopefully the point at which we become hugely profitable.

Phase 9: Other Weird Stuff

Eventually, I don’t want you to even watch the news and be lied to. Or YouTube videos. That is becoming ever more possible. Same with audio media. There’s no reason that couldn’t be all amended under similar principles. What if you could pop on a pair of VR glasses at the grocery store and suddenly you could see all the stuff that’s recommended by people you trust? We will need utilities like this as AI gets more powerful. This could be the backbone of all of that. My full dream for all of this is that you eventually end up in the future where you perfectly trust your politicians to do good stuff for you, and that you’re smart to feel that way.

Okay, So Now What?

A lot of this will depend. Can we attract an investor? How much investment could we get? Will we get all the right people involved? I’ve not ever started a company before but now that I have two kids, I’m back to thinking someone has to do it. The only two things I ever see people seriously suggest online is either “What if we don’t do anything and just ask people to personally perform in-depth investigations every time something seems off to them?” and “What if we have a totalitarian regime where one person or body controls all of information for everyone?” Being able to offload cognitive tasks is the basic building block of civilization.

If someone gave me several millions of dollars, I would quit my job and dedicate myself to this full time, hire the development team, and build the whole thing. That’s the part I’m most confident I could do. The other stuff I’ve not ever done but seems straightforward. I’d start paying out money to contributors to get the flywheel going. I’d probably also build out to phase 4 before going public with adequate funding so I could get explosive growth. Go out and find high conflict media cases, pay for the adjudication to be done by high profile people, advertise our product during the process. That’s something that seems like it would be relatively easy because people are always taking sides on social media and arguing about things, so you might as well take advantage of the eyeballs that the media is going to constantly create for you.

What if I can’t get enough money to do any of that? Well, we tried. The NYT has revenues of a $2.5 billion a year and I’d think a “news entity” that lays across all other news entity probably has a market at least that big, especially if we are getting corporate PR dollars from everyone to refute stuff like “Tuna Fish was invented by Hitler.” There’s a company you’ve not ever heard of called Neustar that is an information broker that’s worth something like $3 billion and it does way less stuff than this.

What if I can convince someone else to build it? Okay. Good. Load off my mind.

What if I get a job somewhere that will let me build something close to it? Even better. The reason I’m so jazzed up about Substack is that it already has all of the right pieces to do something like this very well. Also, nominative determinism. Substack has a name that sounds like full stack subscription service, but really it could be “Below the Stack of Other companies.” It could eat the NYT but in a positive way that causes the NYT to be reborn and for everyone to make more money.

What Can I Do?

Are you a front end developer? Have you ever made a chrome extension? We could use you! Please comment or send me a DM. I can add you to the discord channel and I am working on writing out more requirements for the GitHub we have going.

Are you a person interested in truth with a large audience and a unique perspective? Let me know and we can approach you to be one of our pilot groups.

Are you a person who loves to argue? Good! You can be one of our advocates.

Are you a regular person who just cares about the truth? Good! We will need jurors.

The point of all of this is to make arguments point upward, toward the good and toward truth. We can live in a better future than we currently inhabit and we can live in a systemically better future as well. One that feeds on strength. One that can repel lies.

Please sound off in the comments! And sigh, don’t forget to like and subscribe.