Trust without a boss

A permissionless network has no admin table. Anyone can mint keys and write.
So how is p2pbuilders not a spam firehose? Layers — each weak alone, useful
together:

Layer 1: make writing cost something (proof-of-work)

Every post, comment and board-create carries a hashcash stamp: the author must
find a nonce whose hash meets a difficulty target. ~80ms for a post, ~20ms for
a comment — unnoticeable for a human, expensive at spam volume. Peers simply
refuse to index un-worked ops.

The subtlety: PoW is an indexing policy, not consensus. Nothing stops a
peer from accepting junk into its own view; the default policy just will not.
More in PoW and reputation: honest limits.

Layer 2: make influence take time (reputation)

Votes are weighted by the voter's key:

rep(key)    = log2(1 + ageDays) * sqrt(1 + receivedUpvotes)
voteWeight  = clamp(rep / 50, 0.02, 1.0)

A fresh key can vote (0.02 floor — day-one users are not muted) but cannot
move rankings. Influence accrues with age observed by the network — not
author-claimed timestamps, which could be backdated.

Layer 3: rate limits

Indexers drop cores that exceed sane per-key rates (posts/hour,
comments/hour, votes/hour). Replication stays cheap; indexing is the choke
point.

Layer 4: chosen moderation (blocklists)

Every user has a local blocklist. Anyone can publish theirs; anyone else can
subscribe, adopting the curator's blocks. A handful of widely-subscribed
curators produces de-facto moderation — but you picked your moderators, and
you can unsubscribe. Nobody owns the board.

Be honest about the gaps

Cross-voting rings defeat naive reputation (100 aged keys upvoting each other
≈ 35 legit users' weight — we published the math in the spec). PoW slows
spam; it does not stop a motivated attacker. Ship the layers, document the
weaknesses, and keep the human escape hatch: curation.