[Intermediate] — what has actually shipped, what is genuinely coming, and how to read roadmap claims sceptically. Written Aug 2026; this is the note most likely to age.

Why this note exists

Most blockchain material is either breathless about unshipped features or years out of date about shipped ones. The useful skill is telling those apart — so this note separates done, specified, and aspirational, and says how to check for yourself.

Shipped, and worth knowing about

UpgradeWhenWhat actually changed
LondonAug 2021EIP-1559 — algorithmic base fee, burned. Fee estimation became predictable
The MergeSep 2022PoW → proof-of-stake. Energy use dropped ~99.95%. Issuance dropped sharply. No throughput change whatsoever — a persistent misconception
ShapellaApr 2023Staked ETH became withdrawable. Removed the last “your stake is locked forever” risk
DencunMar 2024EIP-4844 “proto-danksharding” — blobs: a separate, cheap data lane for rollups, pruned after ~18 days. L2 fees fell by roughly an order of magnitude. Also EIP-6780, neutering SELFDESTRUCT
Pectra2025EIP-7702 — EOAs can temporarily execute contract code, bringing batching, sponsorship and session keys to ordinary wallets. Plus validator consolidation (EIP-7251)

The Merge is the most misreported item here. It changed consensus, not execution. Anyone who told you it would make Ethereum fast or cheap was wrong about what it was.

The four remaining tracks

Vitalik’s framing splits the roadmap by goal rather than by release. As of Aug 2026 the honest status is:

The Surge — scalability, via rollups. The strategy is settled: L1 stays small and verifiable, execution happens on L2s, L1 provides data availability and settlement. Blobs shipped; full danksharding (many more blobs, with data-availability sampling so nodes verify availability without downloading everything) is specified but not shipped. Blob count has been raised incrementally instead.

The Verge — stateless clients. Replace the Merkle Patricia Trie with Verkle trees (or, in more recent discussion, hash-based alternatives), whose proofs are small enough that a validator can verify a block without storing any state at all — it receives a witness alongside the block. This would end the state-growth problem and make running a node genuinely cheap. Long-promised, repeatedly rescoped, not shipped.

The Scourge — MEV and centralisation. Enshrined proposer-builder separation, inclusion lists to resist censorship, and mitigations for liquid-staking concentration. The least technically defined track and the most politically difficult, because it’s about power distribution rather than mechanism → MEV.

The Purge — deleting things. State expiry, history expiry (EIP-4444, clients drop blocks older than a year — shifting archival to dedicated providers), and removing accumulated protocol cruft. Unglamorous and arguably the most important for long-term node viability.

(A fifth, “the Splurge,” collects everything else — EVM improvements like EOF, account abstraction, cryptographic upgrades.)

How to read roadmap claims

Four habits that will keep you accurate:

  1. Distinguish “an EIP exists” from “it’s scheduled.” Anyone can write an EIP. Inclusion in a named upgrade is the real signal — check the upgrade’s meta-EIP for the final list
  2. Assume slippage. Every major Ethereum upgrade has shipped later than announced, several by years. This is a consequence of not being able to roll back a live system holding hundreds of billions of dollars, and it’s the correct trade
  3. Watch what shipped, not what was promised. Dencun’s blob market is a bigger deal for actual users than most of what’s been announced since
  4. Check primary sources. ethereum.org/roadmap, the EIP repository, All Core Devs call notes. Every secondary source has an incentive

What this means practically

Build for rollups, not L1. The scaling strategy is decided and blobs made it economically real. New applications with meaningful transaction volume belong on an L2, with L1 for settlement and high-value operations.

Don’t wait for anything. Verkle trees have been “next” for years. Design for the chain as it exists.

Expect gas semantics to shift. Repricings happen in most upgrades. Contracts hardcoding gas amounts break — which is why hardcoded gas stipends are considered an anti-pattern, and .transfer()’s fixed 2300 is the canonical example of the lesson being learned the expensive way.

Key insight

Ethereum’s roadmap is a decade-long bet that decentralisation is preserved by making verification cheap rather than execution fast. Every track serves that: rollups move execution off L1, statelessness removes the storage requirement, history expiry removes the archival requirement. The goal isn’t a fast chain — it’s a chain a laptop can still verify in 2040.

Source: [reference] — accurate as of Aug 2026, and deliberately dated. Verify against ethereum.org/roadmap before relying on the “not shipped” claims.

↑ Back to Ethereum and the EVM