Organise news around the story, not the post — hiddestokvis. Organise news around the story, not the post<br>01 Jul 2026
This is a follow-up to<br>We need a new editor-in-chief.<br>That post made the case that the algorithms have quietly become our<br>editors-in-chief, deciding what we pay attention to, and that even when<br>you tell yourself you don't trust the feed, you trust it the moment you<br>read what it puts in front of you. This post is the other half: not the<br>problem, but a sketch of what a better design might actually look like.
The short version of the diagnosis, in case you skipped the first post:<br>the job of deciding what a society pays attention to used to sit with a<br>relatively diffuse set of editors. It now sits with the few people who<br>own the largest platforms, exercised through feeds tuned for one thing,<br>engagement. I don't think the problem is the algorithms as such. We<br>still have good journalists, and we still read, we just don't read on<br>paper. The problem is the objective function.
What follows is my attempt at a different one. I am not from the media<br>world, so treat this as an outsider's design, offered for people who<br>know the domain to pull apart.
Three structural failures
Most of what feels wrong about social feeds traces back to three design<br>choices, not to malice.
Engagement optimisation rewards the wrong content. Outrage, intensity<br>and confirmation travel further than accuracy or depth, so the system<br>selects for them.
The post is the wrong atomic unit. A subject gets shattered into<br>thousands of disconnected posts, each starting the conversation over,<br>none accumulating into understanding.
Provenance is hidden. You usually cannot tell what is human, what is<br>sourced, or whether the author has ever been right before.
The proposal below is roughly the inverse of these three.<br>The story as the unit
Instead of a feed of posts, you follow a persistent, evolving page about<br>a single subject: a summary, a timeline, sources, open questions and<br>discussion. Think of a Wikipedia article, but time-ordered, so you can<br>see how understanding developed rather than only its current state.
Inside a story, contributions are sorted into four labelled buckets:<br>Reporting: what happened, with sources.<br>Analysis: what it means.<br>Opinion: what someone thinks, kept clearly separate from reporting.<br>Discussion.
Bots are first-class participants but always labelled. They do the rote<br>work, summarising, linking sources, tracking changes, so people can do<br>the judgement.
Mechanisms
The idea lives or dies in the details, so here are the load-bearing<br>ones.
Cross-perspective endorsement. Borrowed from Community Notes: a story<br>only develops when people who usually disagree agree it is worth<br>tracking. This is the main defence against one-sided capture.
Track records over follower counts. Every contributor carries a public<br>record of the claims they have made and how those claims resolved.<br>Standing is earned by being right over time, not by audience size.
Ranking without engagement metrics. Content surfaces on source<br>quality, cross-perspective endorsement and resolution accuracy. Likes<br>and reply volume do not feed ranking at all.
Deliberate friction. To contribute you must cite sources and classify<br>your contribution. This slows the hot take down by design.
Surfaced disagreement. Where framing is contested, parallel summaries<br>and inline annotations show the friction instead of hiding it behind a<br>single authoritative version.
How this differs from things that already exist
The obvious objection is that pieces of this exist already. They do, and<br>that is mostly encouraging.
Community Notes showed that cross-perspective agreement can scale and<br>resist gaming, but it sits on top of an engagement feed and only<br>annotates posts. Here that mechanism is the substrate, not a patch.
Wikipedia has the persistent-article model and a strong sourcing norm,<br>but it aims for a single neutral version and deliberately flattens both<br>the timeline and the disagreement. This keeps both.
The fediverse and Mastodon decentralise hosting and moderation, but<br>they keep the post as the unit and largely keep chronological or<br>engagement-style feeds. Decentralised plumbing, same objective<br>function.
Prediction markets and reputation systems score people on accuracy,<br>which is the track-record idea, but they are narrow and adversarial.<br>The aim here is to fold that scoring into ordinary reading and<br>contributing.
So the bet is not a new primitive. It is combining cross-perspective<br>endorsement, the story as the unit, and visible track records into one<br>system, where each covers the others' weak spots.
Where it breaks<br>I would rather list the holes than pretend there aren't any.
It is structurally conservative. When one side is simply right and the<br>other simply wrong, a consensus-gated system may fail to surface the<br>true story quickly. That is a real cost and I don't have a clean<br>answer.
It diffuses editorial power rather than abolishing it. Someone...