Vitalik Buterin wants a hard cost limit on Ethereum's faster proof generation, and the network is now testing that proposal against exactly the kind of load it was meant to constrain. The core problem: fast sampled blocks don't establish the maximum proving burden a node can take on, and Ethereum's existing CPU gas analysis doesn't count proof generation at all.
The gap in the gas model
Ethereum's CPU gas accounting measures execution work. Proving is a separate job, and it's the expensive one. A block can sail through the gas limit and still leave provers with a bill the current model never sees. That's the hole Buterin's proposed cap is meant to plug — a hard ceiling on proving cost rather than a sampled average that flatters fast blocks.
Sampling is the weak point. Fast sampled blocks show what a node proved on a good run, not the worst case it has to survive. If the cap is calibrated to those samples, it isn't a cap. It's a best-case reading with a limit sign taped to it.
Why the timing matters
Ethereum's push toward faster proof generation is what makes this urgent. Speed and cost pull in opposite directions here. Tighter proving times usually mean more compute per block, and more compute per block is precisely what a hard cost limit would restrict. Testing faster generation without agreeing on the ceiling first means the network finds its limit by hitting it.
The proposal doesn't resolve that tension. It just names it. Buterin's limit sets a number the network has to design around, and the current CPU gas analysis gives builders no way to price the proving work they're actually adding.
What the test is actually measuring
Sampled fast blocks tell you throughput. They don't tell you the maximum proving burden, which is the figure a hard cap needs. Those are different measurements, and conflating them is how a cost limit ends up set too high to protect anyone.
The Ethereum work now in progress is testing faster proof generation against that distinction. Whether the cap holds depends on whether proving cost can be measured directly rather than inferred from fast samples — and on whether the gas model gets updated to include proof generation before the faster path ships.
The open question
No timeline has been attached to the cost limit, and no final number has been set. The next concrete step is measuring proving burden directly instead of through sampled fast blocks. Until that happens, the cap is a proposal running against a gas model that can't yet see the cost it's supposed to cap.
