Irys Blog
TL;DR — Cascade is live on Irys mainnet. Onchain data now has a lifecycle: store it for 30 days, a year, or permanently, priced for the term you choose and instantly downloadable throughout. Cascade also adds partition recovery, a 24-hour commitment-transaction window, a corrected fee baseline, and a set of consensus-hardening fixes.
Operators: upgrade to mainnet-4.0.5 now. Builders: start with the new docs section, What you can build on Irys.
The things we keep
In 1956, an archivist at the National Archives named Theodore Schellenberg published a book that asked a question nobody in his profession wanted to answer. What should we throw away?
For most of the history of archives, the answer had been nothing. The European tradition Schellenberg inherited held that if a document made it into the archive, it stayed. All of it. Forever. But Schellenberg was working in postwar America, where the government was producing paper faster than anyone could shelve it, and he could see where that road ended. If you keep everything, you eventually find nothing. So he proposed a new way of seeing the job. The archivist's real job was not preservation. It was selection. Deciding what deserved to last.
We tend to treat permanence as a virtue. But permanence is also a decision we make by not deciding.
Until today, permanent storage was the only option Irys offered. You wrote something to the network and it stayed, and you paid once for the full price of forever whether you needed forever or not. For some data that is exactly right. Archives, records, and other data meant to outlive the people who made it.
But not everything is worth keeping forever. Most data has a natural lifespan, and yours may be a week or a decade depending on what it is for. Giving you the power to decide is what we have spent this year building.
Cascade is live
Cascade activated on Irys mainnet this morning, the largest change to the protocol since mainnet launch. You can now store data for thirty days, a year, or permanently, and pay for the time you actually need. The 30-day and 1-year options are called term ledgers, and storing data in them costs considerably less than permanent storage. They also change how a dataset behaves over time.
As an example, suppose you keep 100 backup files on a 30-day term ledger. Over the month, you create ten new files, and five of the old ones go out of date. When the term ends, you let those five expire, extend the other 95, and add the ten new ones. You are now paying for 105 files, all of them current, and none of them re-uploaded.
Data expiry happens on its own. Miners earn rewards by continuously proving they still hold the data they are responsible for. When a term ends, those proofs stop counting toward rewards. With nothing left to earn for holding it, miners drop the expired data and free the space. The blocks still keep a permanent record of what was stored and when, even after the data itself is gone.
With most storage providers, paying less means waiting longer to get your data back, because cheap data sits in a slow archive tier. On Irys, term data lives on the same partitions as permanent data and is served through the same gateway, so it stays instantly downloadable for its entire life. The term sets how long your data lives, not how fast you can reach it.
A network that recovers on its own
Consider an undersea cable that fails, cutting every Irys node in Australia off from the rest of the world. The Australian miners keep producing blocks. So does everyone else. Now there are two versions of the chain. When the cable comes back, the network keeps the longer one, the chain with more blocks behind it, and discards the other. Australia, with fewer miners, was building the shorter chain, so its nodes have to give up those blocks and fall back in line. Before Cascade, that meant resetting the node and starting over. Now it rolls back up to 100 blocks, catches up, and keeps running.
On a network built to store data, a reset is expensive. Starting over means downloading all of that data again. Rolling back a hundred blocks instead lets a node recover quickly rather than rebuild from scratch.
Two more changes affect miners directly:
- Commitment transactions, which miners use to register obligations to the network like staking an address or pledging storage, used to expire after 20 blocks. For some operators that was not enough time to sign a transaction and get it posted, so the window is now 24 hours.
- A pricing correction. Block rewards are high early in a network's life, so fees matter less, but over time fees become the main reason to keep storing data. The cost-per-GB baseline now reflects what storing that data physically costs a miner.
Cascade also includes a set of consensus-hardening fixes, detailed in the release notes.
What you can build on Irys
Cascade and the V2 gateway make new kinds of data usable on Irys. The V2 gateway is the upgraded layer that serves Irys data to applications and users, and it now attaches a cryptographic receipt to every piece of data, a proof that what you retrieved is exactly what was uploaded, verifiable in milliseconds. The new docs section walks through three of the things this unlocks in greater detail. These are just examples, not an exhaustive list.
- Chain snapshots. Every blockchain needs a way for new nodes to download their starting state, and today that download typically comes from a third-party server like AWS. On Irys, a blockchain can store its state as data on the chain itself, and any joining node can check every byte against confirmed blocks before it boots. Irys is the first blockchain that can do this.
- Expiring data. Most working data changes over time. New records come in, old ones stop being useful. Until now, onchain data could only be added to, never trimmed. With term storage, a dataset can renew the data it still uses and let the rest expire. Training data, telemetry, and application state all work this way.
- Fully onchain applications. An application's frontend, its data, its smart contracts, and its payments can all run on Irys. The frontend is served straight from the gateway, and the smart contracts run in IrysVM with access to that onchain data. Today even decentralized apps usually rest on a hosted website, and whoever hosts it can take it down. On Irys, the whole application lives on the network.
Read the new docs section: What you can build on Irys.
Each of these makes the next one easier. Cascade opens the network to data that would never pay for permanence, term pricing brings that data onchain, and every dataset that lands becomes shared infrastructure for the next builder. Applications can discover and reuse data that is already there instead of hosting their own copy, and every upload, extension, and retrieval pays fees to the miners who keep the network running. More data brings more builders. More builders bring more data. That flywheel is what a datachain is for.
Cascade is one step toward the future we believe data deserves, onchain or not.
Schellenberg spent his career on a single idea. Keeping everything is not the same as keeping what matters. Seventy years later, Irys is a place where you get to decide which is which.
Get started
- Node operators who have not upgraded to 4.0.5 should upgrade now. Full release notes.
- Builders can start with the new docs section, What you can build on Irys, which walks through term storage, chain snapshots, and fully onchain applications with working examples.
Frequently asked questions
What is the Cascade hardfork? Cascade is a hardfork of the Irys network, activated on mainnet on August 3, 2026 at 12:00 UTC. It adds 30-day and 1-year storage terms at the protocol level, deep reorg support for recovering from network partitions, a 24-hour expiry window for commitment transactions, and a corrected fee baseline.
What are term ledgers? Term ledgers are the part of the Irys protocol that stores data for a fixed term, currently 30 days or 1 year, at a lower cost than permanent storage. Data in a term ledger can be extended before its term ends or allowed to expire.
Is permanent storage still available? Yes. Permanent storage remains the default for data meant to last. Cascade adds 30-day and 1-year terms alongside it, so you choose permanence when you need it and a term when you do not.
What happens when data expires? When data reaches the end of its term, the network stops paying miners to store it, so they drop it and free the space. The blocks keep a permanent record of what was stored and when, even after the data itself is gone. Terms are minimum storage durations, but data can live longer than the term. Expired data stored elsewhere can still be proved against the network at any time.
Is term data slower to retrieve than permanent data? No. Term data is stored the same way permanent data is, on the same partitions, mined by the same miners, and served through the same gateway. It stays immediately downloadable for its entire life.
Does data disappear the moment its term ends? No. Data is guaranteed to be stored and downloadable for at least its full term, and it often stays available for some time after, until miners drop it. Choose a term for how long you need the data guaranteed, not as an exact deletion time.
Do node operators need to do anything? Yes. Upgrade to mainnet-4.0.5 if you have not already. Nodes running pre-Cascade releases no longer follow the canonical chain.
Can other blockchains store data on Irys? Yes. Any blockchain can publish data such as state snapshots to Irys, and its nodes can verify every byte against confirmed blocks before using it. What you can build on Irys covers this in detail.