What Ethereum's Layer 1 Upgrades Actually Change
Layer 2 gets the headlines, but Ethereum keeps rebuilding its own foundation: the Merge, proto-danksharding, statelessness research. What each base-layer upgrade changes, what has shipped, and why the L1 deliberately stays slow.
Rollups get the attention, but they only work because the chain underneath them keeps being rebuilt. "Layer 1 upgrades" are changes to Ethereum's own protocol, as opposed to Layer 2, which processes transactions elsewhere and settles back. The distinction matters because the two evolve on completely different clocks. An L2 can iterate like a startup. Every L1 change has to clear the consensus of a network holding hundreds of billions of dollars. Here is what has actually changed at the base layer, and what is still ahead.
What shipped
The Merge: Proof of Stake (2022)
The largest live infrastructure swap in crypto's history. Ethereum replaced Proof-of-Work mining with Proof-of-Stake validation, mid-flight, without downtime. Energy use fell by about 99.95%, block production became more predictable, and security shifted from hardware to staked ETH. The Merge did not make Ethereum faster. That was never its job. But nothing on the current roadmap works without it.
Proto-danksharding: EIP-4844 (2024)
The upgrade that made the rollup strategy real. It added a new transaction type carrying blobs: chunks of data that rollups use to post their compressed transaction batches. Blobs are priced separately from regular gas and deleted after a few weeks, since they only need to stay available long enough to verify. Rollup costs dropped by an order of magnitude almost overnight, and this is the main reason transacting on an L2 now costs cents. The name is honest: it is the first slice of danksharding, shipped early because it delivered most of the value.
What the plan used to be, and is now
For years the scaling roadmap was execution sharding: split the chain into many shards, each processing its own transactions. That plan is dead. It was enormously complex, and rollups made it unnecessary. If L2s do the executing, the base chain only needs to shard data, which is far simpler. Full danksharding is that endpoint: scale blob capacity until Ethereum is effectively a high-throughput data-availability layer with a settlement court attached, securing an ecosystem of rollups that do the actual work.
What is still research
Statelessness. Every Ethereum node currently drags the full account state around to verify blocks, and that state only grows. Stateless designs let nodes verify blocks using compact cryptographic proofs instead of stored state, which would make running a node dramatically lighter and decentralization cheaper to maintain. Alongside it sit continuous, less glamorous gains: EVM execution optimizations, better transaction propagation, work on faster finality. Individually small, compounding over years.
The deliberate trade
Notice what is not on this list: making the base chain fast. Ethereum has chosen its lane. The L1 stays slow, maximally verifiable and boring, and speed is the job of the L2s. Solana and other monolithic chains make the opposite bet: one fast layer, higher hardware demands, no rollup detour. Which philosophy wins is one of the genuinely open questions in crypto, and it will be decided less by architecture diagrams than by where users and liquidity settle. The companion post on Layer 2 covers the other half of Ethereum's answer.