GHOSTDAG · kHeavyHash · verifiable AI

The same math,
a new frontier

BlockDAG proved proof-of-work can be fast and parallel. kHeavyHash made its core operation the same math that runs AI. Here is what happens when you build on that.

a short read, with things you can poke
Where it starts · GHOSTDAG

Nothing gets thrown away

A classic chain keeps one block at a time and orphans the rest. Every competing block, and the work behind it, is discarded. GHOSTDAG keeps every valid block and orders all of them. A greedy heaviest-cluster rule picks a "blue set" that gives the whole graph a single agreed order, so the network can run many blocks per second without wasting the work that made them.

blue set, kept and ordereda single chain would orphan these
Blocks arrive in parallel. Watch the DAG keep and order them instead of dropping the losers.
The engine · kHeavyHash

Proof-of-work that multiplies matrices

Every mining attempt does three things: a keccak hash, then it multiplies the result through a fixed 64×64 matrix, then keccak again. That matrix multiply is the heavy part, and it was chosen to be deliberately hardware-agnostic. It is also, exactly, the dense linear-algebra operation that dominates how neural networks run. The machines securing a BlockDAG are already speaking the native language of AI.

The same grid, two readings. A proof-of-work hash and a neural-network layer are the same multiply.

The rigs already do the math. The question is what else it can prove.

The turn · SEDRA

A foundation, pointed at a new job

SEDRA is a GHOSTDAG and kHeavyHash network, the same lineage. On top of that base it runs Metacore: a place where an answer is not produced by one model you have to trust, but by several independent ones that cross-check each other, and each answer comes back with a receipt sealed to the chain.

Let's be exact

Mining secures SEDRA. It does not run the AI models, and we will never pretend it does. The AI work is done by people who choose to run models on the network. What mining and AI share here is the foundation: parallel consensus, matrix-multiply proof-of-work, one honest ledger. That shared base is the whole point.

Why it matters · proof

Answers you can check, not just trust

The same instinct that made you care about a chain where work is not wasted applies to AI. On Metacore, every verified answer is cross-checked by a quorum and sealed. Anyone can recompute the seal and check the signature. Here is a real one from the network right now.

loading a live receipt…
Your part

Three honest ways in

No pressure and no promises. SEDRA is an early network, so the numbers are small and the point is the foundation, not a quick return.

Point spare hashrate

Your kHeavyHash rig already speaks SEDRA. Solo mine straight to your own wallet, no fee, no custody. Early network, honest numbers, nothing to leave behind.

Point your rig →

Ask something

Try Metacore. Ask a question or hand it a document, get an answer from an independent quorum, and a receipt you can verify yourself.

Open Metacore →

Run a model

Bring your own model and earn SDR for answers that pass the quorum. The machine that mines can also do this.

Start earning →

Same foundation. New frontier.

SEDRA is not here to replace what you already believe in. It takes the same ideas, parallel work that is never wasted and proof you can check, and points them at verifiable AI.