The Great Battle of Forks

and why we should avoid it

On 6 March 2026, Vitalik published a **post on X** about the potential trade-offs of migrating from Casper FFG to Minimmit. While some of the details are a little too technical, others thoughts and assumptions require only common sense to understand. Therefore, I decided to write an article to start a discussion.

As Vitalik wrote, migrating from Casper FFG to Minimmit means the fault tolerance drops from 33% in Casper FFG to 17% in Minimmit. Further consequences depend on whether the network is synchronous.

Let’s take a closer look at Vitalik’s post. Assume p is the amount of stake owned by an adversary.

In the synchronous case of Minimmit: If p < 17%, nothing bad happens (compared to Casper FFG’s 33%) If 17% < p < 50%, the attacker can prevent finality (incurring losses due to inactivity leakage), but the chain can still progress. If 50% < p < 83%, the attacker can censor or revert the chain, but cannot reverse finality. If p > 83%, the attacker can revert the chain, which is a really bad scenario.

In the asynchronous case of Minimmit, if p < 17%, nothing bad happens (compared to Casper FFG’s 33%). If p > 17%, the attacker can revert finality at the massive cost of slashing.

Vitalik’s arguments can be summarised in three points:

• The network is usually synchronous, so we optimize this work mode,
• The stop finalty attack (17% < p < 50%) is not that bad. The censorship and revert attack (50% < p < 83%) can be mitigated by the ‘minority soft fork’, which involves coordinated developer efforts to create an honest chain that excludes the attacker. All in all, in the usual synchronous mode, the attack can be repelled if an adversary owns less than 83% of the stake (a huge amount!).
• The network may become asynchronous for an extended period, but p > 17% still provides adequate economic security (in the billions of USD) at current prices. Ultimately, however, the network returns to synchronous mode.

However, I disagree with this reasoning and would like to invite the public to verify my thoughts.

There is no doubt that an ‘honest minority’ soft fork can be created, but assuming that the broader user base and capital will follow the “freedom fork” seems overly optimistic. Most people will likely follow the path offering the most liquidity and economic activity rather than the cypherpunk ideals. Ethereum has already undergone one hard fork in response to the DAO hack. The network was split amid the dilemma of the “Code is Law” narrative and the desire to recover the stolen funds. The Recover fork won out economically because it represented the mindset of the broader community and capital. It became today’s Ethereum. Cypherpunks’ Classic Ethereum now has a market capitalization that is 250–300 times lower. I see no reason to assume a different outcome today. People will choose the fork offering the greatest liquidity rather than the greatest freedom. Therefore, the “compliant” fork, supported by all the CEXs and regulated stablecoins that are compliant by default, will win in terms of capitalization — not the “uncensored” one.

I **asked Grok on X **to make a quick estimation of the possible capitalization ratio between forks. Grok assumed that 80–95% of the combined market capitalization will opt for the compliant fork (CEXs, institutions, regulated stablecoins such as USDC, and dependent capital). The “uncensored” fork will account for only 5–20%, respectively. It is reasonable to assume that the price of ether will split accordingly. This would lower the cost of an attack on the “uncensored” fork by a factor of 5–20. Assuming the Ethereum/Classic ratio, the cost of an attack on the ‘uncensored’ fork would be even lower, at 200 times less than the current price.

Is the asynchronous case realistic? I believe it is quite possible if carried out by a state-backed actor, given that internet infrastructure providers can be legally forced to manipulate network traffic. There are multiple examples of state-initiated internet filtering or cooling-off periods around the world right now. Therefore, we should assume the worst-case scenario of an asynchronous attack with prices 200/0,17 ~= 1,200 times lower. This would make the cost of the attacker’s stake around 230 million USD, which is quite achievable. Furthermore, we should assume that an adversary’s expenses will be even lower as they will control the FUD and manipulate the market. Additionally, a state-backed attacker could force dependent actors, such as CEXs, to use their stakes to support the attack. It is also reasonable to assume that the attacker already owns a significant stake and that the remaining gap is not that big. Will the “uncensored“ PoS fork remain sustainable at this lower level of capitalization — and thus security — if it is partially pre-owned by an adversary? I’m not sure.

Furthermore, we should bear in mind that the internet infrastructure and blockchain community were not really designed and battle-tested to survive in an aggressive, hostile environment. Most internet providers are required to comply with regulations. While miners are supposed to be independent economic actors, this only applies in free societies. In highly regulated societies, miners’ priorities (the Miner’s Schelling Point) will shift from validating independent, objective truth and earning money to validating the latest regulations to avoid punishment. Neither network nor community is really ready to survive in the hostile environment.

Splitting the chain into “compliant” and “uncensored” forks is bad and should be avoided rather than encouraged. The “uncensored” part of the network will only survive as an indistinguishable — and therefore inseparable — part of the whole network, not as a separate fork.

Our objective is to build a free network and community on top of an increasingly regulated, less permissive infrastructure and community. In such conditions, plausible deniability and steganography are the only approaches that realistically hold up.

A security model based on one-off forks and slashing is not robust enough. Most capital tends to be compliant and often aligned with state-backed interests, which makes it unreliable as a defensive base. For the system to resist capture, an attacker would need to incur significant slashing losses repeatedly, not just once, while enforcing compliance across the chain. Only then might the losses become too high for the attacker.

But this only works if actors committed to “freedom” must remain active on the chain under attack. They need to force the attacker into the slashing again and again, avoiding escalation into a full-scale “Battle of Forks”. Once non-compliant participants are pushed into a separate fork, the outcome is effectively decided: the attacker retains the original chain with the most capital on it and can progressively tighten compliance constraints without any resistance.

What do you think?

P.S. Many thanks to Brian Barrett, who reached out to me because of my post on X and encouraged me to write this article.