Blob data is stretched with extra redundancy so that any half of it can rebuild the whole, like a puzzle you can finish with half the pieces.
The idea: Treat the data as points on a polynomial curve, then evaluate the same curve at as many extra points again. Any half of the points pins down the curve, so any half of the extended data rebuilds all of it. A block producer who wants to hide anything must withhold more than half, so a node that checks a few random pieces will almost surely hit a missing one.
Why it matters: Nodes can check data is available by sampling a few pieces instead of downloading everything.
Stretching data so any half rebuilds it
The 4 data values (blue) are treated as points on the simplest curve through them, which is then sampled at 4 more places (green). Any 4 of the 8 points pin down the same curve, so even with half of them missing (crossed out) the curve, and with it every original value, can be rebuilt. Drawn here with ordinary numbers. The real thing uses remainder arithmetic, and Ethereum's blobs use the same idea with thousands of points.
The details
On the map: PeerDAS: any half of the data rebuilds the blob
Before blob data is distributed, it is extended with a Reed-Solomon erasure code to twice its size[1, 2], the same family of codes that lets a scratched CD still play. Any half of the extended data is enough to rebuild the whole blob.
This is what makes data availability sampling work: if a block producer hides more than half the data, random samples catch it with high probability[3], and if at least half is available, everyone can reconstruct the rest. Since Fusaka (December 2025), nodes only download and custody a subset of columns (PeerDAS) instead of whole blobs[4, 5].
The coding itself needs no cryptographic assumption. KZG commitments are what prove the extension was done correctly.
Rests on
If any of these breaks, this breaks too.
Good randomness(needed by everything in the unconditional box (and everything else))
If it falls
If this breaks, these break with it.
Nothing on the map depends on it: it is an end point, something people use directly.
References
I. S. Reed and G. Solomon. Polynomial Codes Over Certain Finite Fields. Journal of the Society for Industrial and Applied Mathematics, 1960. doi:10.1137/0108018
Mustafa Al-Bassam, Alberto Sonnino and Vitalik Buterin. Fraud and Data Availability Proofs: Maximising Light Client Security and Scaling Blockchains with Dishonest Majorities. arXiv:1809.09044, 2018. arxiv.org/abs/1809.09044
Danny Ryan, Dankrad Feist, Francesco D'Amato et al. EIP-7594: PeerDAS - Peer Data Availability Sampling. Ethereum Improvement Proposals (Final), 2024. eips.ethereum.org/EIPS/eip-7594