Ethereum still finalises in 12.8 minutes. Two epochs of attestations, 64 slots, and only then does the network agree that a block is irreversible. In 2026 that is the same finality budget Ethereum had on Merge day in September 2022. Pectra didn't change it. Fusaka won't change it. The upgrades that change it are further out — and they are the upgrades that will define what Ethereum looks like in 2030.
This is the closing piece of our four-part series on Ethereum's roadmap. Parts 1, 2, and 3 walked through what shipped and what's queued for the next 18 months. This part is about the long tail: the three upgrades that are technically hardest, politically most contested, and — if they land — most transformative. Single-slot finality (now reframed as 3-slot finality), Verkle trees, and the Ethereum Object Format. None of them is close. All of them matter.
Why 12.8 Minutes Is a Problem
Finality is the moment a block becomes irreversible without a hard fork or a slashing event. Today, that takes two epochs — roughly 12.8 minutes — because Ethereum's Casper FFG layer needs two rounds of validator votes to lock a block in. For most users this is invisible. Exchanges wait 64 confirmations anyway. Bridges wait longer. But under the hood, 12.8 minutes is a long time for a chain that wants to be the settlement layer for the world.
It creates real problems. MEV searchers exploit the gap between proposal and finality. Reorgs, while rare, are still possible inside an epoch. Light clients have to trust sync committees rather than direct finality proofs. And bridges to other chains have to assume the worst case, which means slower withdrawals and higher capital lockups.
The fix has a name: single-slot finality. Vote, aggregate, finalise — all inside one 12-second slot. Vitalik first sketched it in his Endgame essay in December 2021, and it has been the polestar of the consensus roadmap ever since.
The Brute-Force Problem
The trouble is that brute-force SSF doesn't scale. Ethereum has roughly one million active validators in 2026. Aggregating one million BLS signatures every 12 seconds — and propagating the result across the global validator set in time for the next slot — is, by current estimates, infeasible. The bandwidth and CPU budget on a typical home staking box can't sustain it.
Researchers spent three years looking for a way around this. The answer that emerged through 2025 wasn't a clever signature scheme. It was a reframe. In his 'Simplifying the L1' post from May 2025, Vitalik proposed splitting the vote across three slots instead of cramming it into one. Slot one proposes. Slot two attests. Slot three finalises. Total wall-clock time: about 36 seconds. Beam call #6 on May 16, 2025 formalised this as 3-Slot Finality, or 3SF, and the consensus research team adopted it as the working SSF candidate.
36 seconds is not 12 seconds. But it is not 12.8 minutes either. It is a 95 percent reduction in finality latency, and it fits inside the bandwidth envelope of a normal validator. The Ethereum Foundation's February 2026 roadmap targets 3SF for end-2026. That target is aspirational. Realistically, given how Ethereum upgrades have slipped historically, expect 2027 or 2028.
What 3SF Changes for Stakers
If 3SF ships, it almost certainly ships alongside a lower staking minimum. The current proposal is 1 ETH per validator instead of 32. The reasoning is that with faster finality and more efficient aggregation, you no longer need 32-ETH chunks to keep the validator set manageable.
A 1-ETH minimum would dramatically expand the staker base. It would also compress yields. With more ETH staked competing for the same issuance budget, APRs trend down — current models suggest 2 to 2.5 percent, versus the 3.0 to 3.2 percent solo stakers earn today. Liquid staking tokens and pooled providers would still take their cut on top.
We covered the staking economics in detail in our staking and LST guide. 3SF is the upgrade that finally makes solo home staking accessible to people with one ETH instead of thirty-two — but it also makes solo staking less lucrative in absolute terms. That trade-off is intentional.
Verkle Trees: The State Problem
The second long-tail upgrade is Verkle trees. Ethereum's state — every account balance, every contract storage slot, every nonce — is stored in a Merkle Patricia Trie. To prove that a particular account exists with a particular balance, you generate a Merkle proof. In the worst case, those proofs can be 18 megabytes for a single block's worth of state access.
That's a problem because stateless clients — clients that don't store the full state, only verify proofs — need those proofs to be small enough to gossip across the network in real time. 18 MB is too big. Verkle trees fix this by replacing the Merkle structure with vector commitments based on KZG polynomial commitments. The math is dense, but the result is simple: proofs shrink from megabytes to roughly 200 bytes per account.
Verkle has been on the roadmap since 2021. It was supposed to ship with Pectra. It got pushed to Fusaka. It got pushed to Glamsterdam. It is now tentatively scheduled for Hegota in the second half of 2026, which realistically means 2027 or later. Each delay reflects the same problem: Verkle is genuinely hard to implement correctly, and it requires a coordinated migration of the entire state tree.
When Verkle does land, it unlocks stateless clients, which unlocks light clients that don't trust sync committees, which unlocks proper trustless wallets and bridges. It is foundational. It is also late.
EOF: The EVM Cleanup That Won't Stay Dead
The third long-tail upgrade is the Ethereum Object Format. EOF is a structural overhaul of EVM bytecode. Today, when you deploy a contract, the EVM has to validate the code every single time it runs — checking jump destinations, stack depth, opcode validity. EOF moves that validation to deploy time. Code is wrapped in a magic prefix (0xEF00), validated once, and then executed without runtime checks.
The payoff is real: cheaper execution, safer code (entire classes of bugs become unreachable), and a cleaner foundation for future EVM upgrades. The anchor EIP is 3540, with 3670, 4200, 5450, 7620, and 7873 building on top. Understanding the EVM and account model helps here — EOF is essentially a version 2 of the bytecode format, and version 1 contracts continue to work alongside it.
EOF was scheduled for Fusaka. On April 28, 2025, after sustained community pushback over complexity and tooling readiness, it was removed. Some developers argued the cleanup wasn't worth the migration cost. Others argued the format was over-engineered. The result was a delay with no clear new target. EOF is still on the roadmap. It is not on a hard fork.
The Endgame Vitalik Sketched
Step back from the individual upgrades and the shape of the endgame becomes clearer. In December 2021, Vitalik published the Endgame essay. The thesis: block production will inevitably centralise, because building the most profitable blocks requires sophisticated MEV extraction and high-performance hardware. That is fine, the argument goes, as long as block validation stays decentralised, censorship resistance is enforced via inclusion lists, and the entire L1 can eventually be verified via succinct ZK validity proofs.
3SF, Verkle, and EOF are pieces of that endgame. 3SF makes the validator set scalable enough that anyone can validate. Verkle makes proofs small enough that anyone can run a stateless client. EOF cleans up the execution layer enough that it can be ZK-proven efficiently. The pieces interlock.
None of this changes the supply story in the short term. ETH issuance in May 2026 is still around 1,700 ETH per day, with 50 to 150 ETH burned per day depending on activity. Net inflation is mildly positive at around 0.23 percent per year, with total supply at 120.7 to 121.5 million. The endgame upgrades don't change the monetary policy. They change what the chain can do. If you hold ETH long-term, you can self-custody it in Zelcore while these upgrades unfold over the next four to six years.
Series Close: From Pectra to Endgame
This series traced Ethereum's roadmap across four upgrades and roughly five years of development.
Part 1 covered Pectra and Fusaka — the upgrades that shipped in 2025 and early 2026, raising the validator stake cap to 2,048 ETH and laying groundwork for blob scaling.
Part 2 covered PeerDAS and the data availability problem — the upgrade that made L2 fees roughly 8 to 10x cheaper by letting nodes sample blob data instead of downloading all of it.
Part 3 covered Glamsterdam: ePBS, BALs, and the MEV redesign — the in-flight upgrade that separates block proposers from block builders and makes MEV extraction more transparent.
This part covered what's still ahead: 3-slot finality cutting finality from 12.8 minutes to 36 seconds, Verkle trees enabling stateless clients, and EOF cleaning up the EVM. None of these are close. All of them are in motion.
The pattern that emerges across all four parts is the same. Ethereum upgrades slip. They get reframed. They split into smaller pieces. And then, eventually, they ship. Pectra slipped from 2024 to 2025. Fusaka lost EOF. Glamsterdam swapped FOCIL for BALs. SSF became 3SF. The roadmap is not a schedule. It is a direction. If you're holding ETH or building on it, that direction is what matters — and the direction, four years after the Merge, is still toward the chain Vitalik sketched in 2021.
