Seals a whole blob of rollup data into one tiny value, with tiny proofs for any piece.
The idea: Write the data as a polynomial: a curve passing through all your values. The setup hides a secret point τ inside curve points, so you can compute "the polynomial at τ", hidden, without knowing τ. That one point is the commitment. To prove the value at some position is y, you show that the polynomial minus y divides cleanly by (x - position), which takes one extra point that a pairing can check. Since nobody knows τ, nobody can find a different polynomial matching the same commitment.
Why it matters: Makes cheap rollup data (blobs) possible.
On the map: EIP-4844 blobs, PeerDAS cell proofs, point-evaluation precompile
A KZG commitment fixes a whole blob of data (a polynomial) with one 48-byte value[1], and short proofs show what the data contains at any point. Since EIP-4844 (Dencun, March 2024)[2, 3] every blob carries one, and PeerDAS uses KZG cell proofs[4] so nodes can check individual pieces they sample.
Contracts verify KZG openings with the point-evaluation precompile[2], which is how rollups prove what their blobs contained. KZG needs the 2023 trusted setup and the q-SDH assumption. The post-quantum plan (leanDA)[5] replaces it with hash-based commitments.
If any of these breaks, this breaks too.
If this breaks, these break with it.