Was ist ein DAG?
Keryx, GHOSTDAG und das Ende der Block-Warteschlange

The picture in your head is wrong (a little)

Ask someone to draw a blockchain on a serviette and they will draw a queue. One block, then another block, then another block. Polite, orderly, single file. Very British. The longer the queue gets, the more secure it is, and everyone agrees on whose turn it is to mine the next one. Lovely.

Now imagine ten people simultaneously decided it was their turn. In a traditional chain, the network looks faintly embarrassed, picks one of them, and pretends the others didn’t happen. The discarded blocks are called orphans, which is unkind, and they take their hashing work with them into a sort of digital wilderness. We are then asked to call this fair.

A DAG — Directed Acyclic Graph, if we’re being formal — politely declines this arrangement. It says: let’s keep all of them, thank you.

What does “DAG” actually stand for?

It is, as the name suggests, three words doing rather a lot of work:

In a regular blockchain, each block has exactly one parent. In a BlockDAG, each block has as many parents as it cared to acknowledge — usually all the recent blocks the miner could see when they started hashing. The result is less of a queue and more of a small, well-mannered party where everyone keeps a mental list of who arrived in what order.

Linear chain vs. DAG, side by side

If diagrams help — and the engineers responsible for this stuff would tell you they very much do — here is the difference in the two scenarios miners find themselves in:

[A] → [B] → [C] → [D]         
                  → [C′] (orphan, sorry)
                  → [C″] (also orphan, sorry)

       ↗ [C ] ↘
[A]→[B] → [C′] → [D] → ...
       ↘ [C″] ↗

Same Proof-of-Work. Same miners. Same electricity bill. Vastly different fate for the blocks that arrived within a whisker of each other.

Enter Keryx, stage left

Keryx — ticker KRX, algorithm keryx-hash, organised by the unhurried sorts at Keryx Labs — is one of the cleanest current examples of a BlockDAG in the wild. Its consensus protocol is called GHOSTDAG, which is (a) cooler than “Optimised Block Inclusion Greedy Sub-Tournament,” and (b) what it actually does.

The numbers that make people raise an eyebrow:

Ten blocks a second is the bit that startles people. Bitcoin gives you a block every ten minutes; Ethereum, every twelve seconds; Keryx, every tenth of a second. If a traditional chain had tried that without GHOSTDAG, it would have spent most of its life orphaning blocks faster than miners could produce them, which is, as is traditional in these matters, not ideal.

What GHOSTDAG actually does (the genuinely clever bit)

Keeping all the blocks is the easy part. Agreeing on the order they should be considered in — which transaction comes first, which double-spend wins — is where most attempts have, historically, fallen on their faces.

GHOSTDAG’s solution is brisk and surprisingly readable:

The polite fiction here is that “blue” means “honest, well-connected, behaved themselves” and “red” means “slightly late to the meeting.” The protocol doesn’t need to know anyone’s motives — it just needs to know who referenced whom. From that, it produces a single, deterministic total ordering of every transaction the network has ever processed. Everyone arrives at the same answer. Nobody throws blocks in the bin.

If you’d like the formal version, the original paper is “PHANTOM and GHOSTDAG: A Scalable Generalization of Nakamoto Consensus” by Sompolinsky, Wyborski and Zohar. It is, if we’re being honest, dense as a Christmas pudding. But the idea above is essentially it.

What this means if you’re the one pointing a rig at it

You may be wondering, somewhere around now, whether mining a DAG involves doing anything different. Short answer: no. Slightly longer answer: no, but the rhythm changes.

Your miner does what miners have always done. The pool sends you a job. You grind hashes. You submit shares when one of them clears the share target. The pool handles the small matter of telling Keryx about every block your share contributed to and which parents to reference. You don’t need to know about blue sets, red sets, or topological ordering. You just need a build of your miner that supports keryx-hash.

What does change is what you see on the pool dashboard:

That last row is worth pausing on. A DAG chain throws fresh jobs at your miner constantly — not because anything is wrong, but because there are simply more blocks. Mining software designed for ten-minute Bitcoin can choke on this if it wasn’t built with brisk pacing in mind. Use a build that knows what it’s doing.

Where it almost goes wrong (and doesn’t)

If you’ve read this far you may be suspicious. “Hang on,” you might say, “if any miner can reference any parent they like, what stops someone deliberately ignoring honest blocks and trying to substitute their own?” This is the British way to disagree: politely, but with the suspicion that someone is up to no good.

The short answer is nothing stops them trying, but GHOSTDAG’s blue/red split punishes them for it. A block that ignores the well-referenced majority gets categorised as red. Red blocks still earn their reward — nobody is left out of pocket — but they sit at the bottom of the ordering, and any double-spend they tried to slip through arrives after the honest transaction did. From the chain’s perspective the attacker simply wasted electricity to be told off.

This is also why a 51% attack on a DAG isn’t cheaper than on a linear chain. Security still rests on total Proof-of-Work in the blue set. If you don’t have more honest hashing than the rest of the network put together, you don’t get to define what blue means. Pool-side security matters in the usual ways; the DAG itself doesn’t weaken any of them.

So — should I point a rig at it?

The honest answer, as ever, is: it depends. KRX is a young chain. Reward, difficulty and price all move quickly while a network is young, and as we have explained at length elsewhere, picking what to mine is mostly a question of taste, electricity costs, and how long you intend to leave the heater on.

What we’ll happily say is this: a BlockDAG is genuinely a different shape of consensus, not a marketing repaint of the old one. The economics, the latency profile, and the feel of mining all shift slightly. If you’ve been mining one Equihash variant for the last four years, an afternoon spent watching a Keryx dashboard tick over is genuinely informative. And if you want to have a go, we run a pool, the fee is 0% during the launch period, and the stratum is at krx.suprnova.cc. No biscuits required.

Frequently asked, briefly answered