7
findings running through the whole reportgrowing from Chapter 02 all the way to Chapter 12
6
directions for policy advocacynot intervening in state processes・producing directions only
100%
of the spam the invitation-only “Chit-chat” channel cleared out in half a monthcommunity design is cheaper and more effective than a model

The seven findings that run through the whole site

  1. The real power of governance lies in “deciding who gets seen”; deleting content is secondary.
  2. Tools grow out of the ecosystem; not a single one was designed out of thin air.
  3. Every disposal step lowers the cost of “not being able to undo a mistake.”
  4. The cheapest and most effective governance is good community design, not necessarily a stronger model.
  5. On “account-ism vs. work-ism,” no one can yet say who is right.
  6. The domain-query rewrite (DNS-RPZ) incident showed us: the same technology can both block harmful content and be used to shut down legitimate people.
  7. Layered governance lets censorship resistance and censorship coexist: the second layer preserves the content, the third layer restricts its visibility.
The cheapest and most effective governance is not necessarily a stronger model. A neighborhood watch team clearing hundreds of spam posts in a week is cheaper — and more effective — than a model.

Policy recommendations (not intervening in existing state policy processes; producing directions for advocacy only)

  1. Proportionality: an individual violation at the content layer should not readily lead to a whole-site block at the infrastructure layer.
  2. Cross-agency reporting and evidence-format standards, so that misjudgments can be corrected in real time.
  3. Remedies must be “actually usable,” not merely “existent”: they have to be available even over a long holiday weekend.
  4. Reducing visibility also requires due process: visibility restrictions should follow Article 17 of the EU Digital Services Act (DSA), addressing notice and appeal.
  5. Layered disclosure, balancing safety and transparency: platforms should not be required to proactively and comprehensively monitor every piece of content.
  6. Multi-stakeholder governance, so that the value judgments behind the thresholds are not decided unilaterally by engineers.
The world cannot consist only of big platforms; many meaningful, pluralistic experiments happen at the edges, in local and particular places.

What this report documents is the real process of a small platform muddling through by trial and error over five-plus years — failing again and again, yet always adjusting — not a perfect system. What makes it truly rare is its willingness to lay the entire process open for everyone to see. Long live Matters.

“The best platform governance is to make the process public, laid out in the sunlight; it's just that in an age of bots endlessly flooding the feed, what people can do is often only to hold the line as best they can in the cracks.”

Further reading: overview of all long-form pieces