lcn_8888 on Nostr: *glm 5.3* 1. Weak subjectivity: your correction is basically right Your understanding ...

*glm 5.3*

1. Weak subjectivity: your correction is basically right
Your understanding matches the spec: it's not that others need to be "online as often as you," but that you must sync at least once per Weak Subjectivity Period. That period is calculated so that any attacker who controls the validator set at the start of the period loses at least a threshold amount of stake (a fraction of , with a SAFETY_DECAY margin) if they finalize a conflicting Checkpoint ethereum.github.io.

Two nuances:

It's not that "the overwhelming majority of stake remains locked," but that the weak subjectivity period is kept shorter than the validator withdrawal period, so that anyone who double-signed cannot withdraw their stake before the slashing evidence is included on-chain ethereum.org. It's a condition about withdrawals vs. slashing, not about how much total stake is locked.
The guarantee is economic and accountable, not one of impossibility: a sufficiently rich attacker can still execute a long-range attack or a live one — it just costs them burning ~1/3 of the total stake (billions of dollars). "Billions must have been slashable" is exactly Casper's accountable safety property, so you're correct there.
And yes: after 20 years offline, the procedure is to obtain a recent checkpoint from independent sources and cross-check them ethereum.org. Your analogy with being able to "hear" the true Bitcoin chain is reasonable: the difference is one of degree, not of nature.

2. ZK proofs: you're right, with one precision
Correct: a zk proof of consensus demonstrates that a chain is internally valid, not that it is the canonical chain. The subjectivity of chain selection simply shifts to whoever chose the proof's input. Demanding raw signatures (which is what a normal client sync does anyway: checkpoint + forward signature verification) does protect you ethereum.org.

One minor precision: raw signatures by themselves also don't tell you which chain is the real one — the defense is the combination of checkpoint + signatures, because a finalized conflicting chain implies slashable double-signatures. The point of the zk system was to remove accountability (hiding who signed); refusing to accept it closes that vector.

3. Anchoring to Bitcoin with OTS: a good idea, but with a conceptual flaw
The idea is genuinely good and has been discussed (PoS chain checkpoints anchored to Bitcoin), but you're misidentifying where the trust ends up:

What it does achieve: it turns "ask a friend or a block explorer for a checkpoint" into a historical, immutable, publicly auditable commitment. A long-range attacker trying to sell you a fake history would have to present state roots inconsistent with the weekly anchors, or have compromised the anchoring itself. That defeats the classic long-range attack.

The flaw: you don't remove the trusted party, you relocate it. Someone has to decide which root gets anchored each week, and you, in 20 years, cannot verify from Bitcoin that that root corresponded to the real chain (verifying that requires... an Ethereum node; it's circular). OTS only proves when a hash was committed, not that the hash is honest. Security then rests on the honesty of the anchorer(s) — ironically, the same type of assumption you're criticizing in Rootstock's pegnatories.

Additionally:

"The attacker would have had to start 20 years ago and run a parallel chain" isn't quite right: if they corrupt or are the anchorer, they need none of that. Real-time effort only becomes necessary if the anchorer is honest and you demand raw signatures.
It's better to anchor finalized beacon block roots, not just execution-layer state roots: a state root commits the state, not the history, and technically there could be alternative histories that converge to the same state at each anchoring point.
The natural mitigation is the same one ethereum.org recommends for checkpoints: many independent anchorers you can cross-check, not just one ethereum.org.

That said, as an improvement over the status quo (ephemeral sources that may vanish in 20 years) it's real: a history of commitments on Bitcoin is more durable and auditable than a block explorer.