The bytes of the first blobs as pixels in a grid

Blobs Are Temporary and Permanent

On the beauty of cryptographic evanescence; and other curious tidbits from the first few weeks of blobs

Takens Theorem
Published in
6 min readApr 9, 2024


The latest Ethereum hard fork, known as “Dencun,” was deployed around March 13th — at epoch 269,568, to be precise. Thousands watched a livestream as this threshold arrived, accompanied of course by a Jonathan Mann tune to celebrate the upgrade, and a captivating animation “Delivery at Dawn” to support Ethereum core developers. This animation was collected on Zora as an open mint, and raised over 170 ETH for developers.

Frame from “Delivery at Dawn,” see here for full video

The celebratory spirit was well placed: Ethereum core developers pulled off another major mainnet upgrade without mishap. The famous philosophical parable “Neurath’s boat” describes science as like a ship at sea that must be repaired and upgraded while navigating. Like the sailors of Neurath’s boat, Ethereum’s core developers add major components and upgrades while mainnet’s blocks tick away without interruption.

Dencun was a major upgrade, combining two sets of updates, Deneb and Cancun, and it included several important Ethereum Improvement Proposals (EIP), such as making some information about Beacon consensus available on the main chain (EIP-4788). The most heralded was EIP-4844, known as proto-danksharding: data blobs. After epoch 269,568 arrived, blobs began dropping into the consensus chain, on Beacon. We summarized blobs in a prior post that anticipated this upgrade. Check out lots of details here.

Dune wizard 0xRob shows blob cost on 4/7 vastly lower than calldata

The goal of blobs was to reduce the cost of operating second layers by creating a new fee market for temporary data. Blobs are temporarily stored on the Beacon chain for 18 days and have a demand-based cost function distinct from mainnet’s gas fees. Once Dencun went live, the effect of the blobs was almost instant. Data scientist and creator 0xKofi created a site to track fees across second layers, called This site showed that when second layers integrated blobs their fees dropped precipitously. At the time of this writing, an ETH transfer on Arbitrum, Optimism, Zora and Base costs less than $0.01.


L2 Starknet showed a surprising number of blobs needed to maintain its ledger, a dominant source of blobs early after Dencun. This is shown below in Dune wizard hildobby’s analysis. Why did Starknet produce so many blobs? At first, its ledger required almost 2 blobs per L2 block. They modified their use of blobs, and you can see a few days later Starknet’s usage drops considerably. This illustrates an intriguing detail of Dencun and EIP-4844: L2 chains took on distinct strategies at deployment, some using blobs immediately and adjusting like Starknet, and others waiting to deploy (for governance decisions, development, etc.), like

A couple of weeks after these initial events, inscriptions arrived in force. As anticipated by Anthony Sassano and discussed in our prior post, many users looked to blobs as a new way to encode inscriptions on Ethereum. As we described in our prior post: One could imagine the blob fee market for L2s being consumed by such playfulness in a manner similar to the rise of Bitcoin ordinals and its accompanying debates. Blobs could be a new canvas; their evanescence held at a premium: “I owned that for those 18 days!”

Dune wizard hildobby shows the spike in inscriptions across March and April

With a website by the creators of Ethscriptions, users can encode images or other data onto blobs. Developers introduced a standard that links a calldata inscription with a Beacon blobscription they called an “Ethscription attachment.” This tool became wildly popular in the two weeks after Dencun. Indeed, “blobscriptions” rapidly took over blobs about a week ago, though their popularity has waned in the past few days.

BlobScriptions landing page to inscribe on blob data

Blob data is temporary, though many services (like The Graph and Etherscan) are storing blob data for the foreseeable future. An elegant feature of blobs is that, although their data expires, provability of a given chunk of data does not. Because Beacon chain preserves cryptographic features of the blob, known as the KZG commitment, it is possible to prove that a given batch of data belongs in a given transaction. In other words, as long as someone has the relevant data stored somewhere, the “truth” of a particular blob of data can be proven — even if Beacon clients have pruned them after 18 days. Put playfully, blobs are temporary and permanent.

View 10,000 recent blobs on Etherscan
This blob data contains a bunny

Was the spike of blobscription activity due to a misunderstanding, perhaps inscribers did not know that blobs expire? Some speculated that this evanescence is attractive in itself, and burning ETH to inscribe blobs is certainly welcome in a culture known for its profligate spending habits.

Blobscriptions have not stopped though. They continue, albeit at a much lower proportion of overall blob activity. Maybe the weeklong spike was the usual momentary mania, generating increased fees, finding intrinsic value to blobscriptions — and it’s now settled back into a simmer.

So another possibility is that the blob fee market operated as planned. This fee market is a demand-based dynamic that sets the cost per byte to store blobs. There was a significant spike in the cost to commit blobs across that week of blobscriptions. The market then rebalanced.

Data from Blobscan; y-axis is bob gas price in gwei per blob gas

Like surplus fees on mainnet, surplus blob fees are burned, too. And so it’s also possible that some users of the chain were delighted in commemorating Dencun with some inscriptions and, as a result, collectively bumped up the ETH fees burned. The “ultrasound money” meme is rooted in an enthusiasm for this monetary policy. Over 700 ETH has been burned through blob fees.

Gas burn visuals, calculations by

Whatever the reasons for inscriptions, blobs are in themselves a beautiful addendum to the chain. Each blob is a mere 128KB. Each sits delicately and modestly for a fruitfly’s lifespan of a mere 18 days. But their simplicity contributes a new dynamic on chain. They have massively impacted L2’s, dropping their fees by orders of magnitude in some cases, even when things get busy on the second layer. They are a canvas for inscriptions, playfully temporary yet permanent in their cryptographic proof characteristics.

Further Reading

I’m on X. I create things and write. Disclosure: I was not paid for this post, and sometimes I own stuff I mention. I wrote this post for fun, and thank the folks at Etherscan for the encouragement and support.



Takens Theorem

Dynamic distributed data displays. Intermittent. Friendly.