FORKED.gg ← All news
← Back to News
Under the hood · 30 August 2026

Is it actually decentralized, or do you just pin an IP?

Someone in our Discord asked the version of this question that deserves a real answer. Not the slogan, the skeptical one: is the compute really spread across many machines, or do you pin an IP address, put it on a list, and call the list decentralized. Fair question. Here is the whole thing, drawn out piece by piece.

A chunky voxel diorama in a glowing green server room: the Forky mascot, a silver fork with a green scarf, and a happy gingerbread-man Forkie hold an architecture blueprint between them, surrounded by server racks, floating green node cubes, a glowing data path on the floor, and a friendly robot and dragon watching from behind

The question came in from our Discord, word for word:

Is there a public github? Curious which parts are decentralized. Like is the actual compute decentralized or do u just pin an ip address and that list is decentralized.

GingerbreadMan, in the Forked.gg Discord

It is exactly the question you should ask any project that uses the word decentralized. Plenty of them mean a spreadsheet of servers with a nice map behind it. So here is how ours actually works, with no hand-waving, and a couple of admissions along the way that most projects would rather leave out.

Start with one request

Forget the architecture words for a second and just follow a single request. You are on your own machine. You want something, a game server, a file, a stream. Watch it travel all the way there and back.

Four numbered steps: 1, your own box asks the coordinator for what you want; 2, the coordinator finds a node that can serve you and hands it a signed work order; 3, an independent node does the real work and produces a proof; 4, the result streams straight back to your box and the proof goes back so the node gets paid only if it checks out
One request, start to finish. The bytes travel node to you directly, not through the coordinator. Open it full size

Four steps. You ask. The coordinator finds a machine that can serve you and hands it a signed work order. That machine, run by somebody who is not us, does the work and produces a proof that it did it correctly. The result comes straight back to your device.

The part worth slowing down on is that last hop. The bytes go from the node to you directly. They do not route back through us on the way. The coordinator points you at the right machine and keeps the tally of who did what, and then it gets out of the way. There is no single box quietly serving everything from behind a pinned address. The work sits out on the machines, and the traffic runs straight from whichever machine holds your thing to you.

node → youcontent is served straight from the machine that holds it to your device, not relayed through a central server
4 proofsstorage, availability, metering and uptime, each one catching a different way a machine could lie about its work
0 earnedwhat a node collects for work it cannot prove it did, no reduced rate, nothing at all

What actually runs out there

The things you can point at and use all run on that same network of machines.

Four products that run on the network: game hosting spins up a live game server on demand, CDN and files are served straight off the nodes that hold them, Forked Play streams uploaded media back, and FORKcast is the prediction engine
Different products, one network underneath.

Game hosting spins up a live server on demand. The delivery network serves files straight off the nodes that hold them. Forked Play takes media you upload and streams it back to you. FORKcast is the prediction engine. Four different products, and every one of them is doing real work on machines spread around the world rather than on a rack in a room we rent.

The one central piece, and we will say so

Here is the first admission, because leaving it out would make this whole post the kind of thing it is arguing against.

The coordinator, labelled the only central piece today: it assigns each job to nodes that can do it, issues signed work orders the node must prove it followed, routes traffic straight to the node holding your file, and keeps the hash-chained ledger of who did what work
One coordinator today. It hands out the work. It does not get to invent the results.

There is one coordinator today. It assigns each job, it issues the signed work orders, it routes you to the right machine, and it keeps the ledger. That makes this a coordinated network backed by cryptographic proofs rather than a fully permissionless one. If someone tells you Forked is already trustless from end to end, they are ahead of where the code actually is. The pieces that push past a single coordinator, a peer-run challenge system and on-chain governance, are built and have already run on testnet, and that is the direction of travel. Today the coordinator is the one central piece, and we would rather you hear that from us than catch us on it.

What the coordinator cannot do is fake the results. It hands out the work. It does not get to write the proof that the work was done. That one distinction is the whole design, and the rest of this post is really just an explanation of it.

The network itself

A grid of independent nodes, some cloud machines and some home machines. Independent operators each run one binary on their own hardware; they are the ones that store the file pieces, serve the bytes, host the game servers and answer the requests. Every node carries its own cryptographic identity and an owner.
Independent operators, one binary each, their own hardware.

The machines are run by independent operators. Each one installs a single program on hardware they own, a cloud box or a computer sitting at home, and that program is what stores the file pieces, serves the bytes, hosts the game servers and answers the requests. Every machine carries its own cryptographic identity and a named owner, so a node is a specific accountable thing rather than an anonymous line on a list. When the license gate switches on, joining the network also takes an on-chain license, so the set of machines becomes something you can check for yourself instead of something we ask you to believe.

Some of the engineering that makes this work, the node layer and the way work is spread and served across it, is patent-pending. We will not walk through the guts of it here, but it is ours and it is on the record.

Why you do not have to trust any of it

This is where the pinned-IP worry actually gets answered. A list of addresses means very little on its own, because an address does not prove that anyone behind it is doing anything. So nothing on this network gets paid for saying it did the work. It gets paid for proving it.

Four independent proofs. Proof of storage challenges a node on the real bytes it holds, a Merkle check not its word. Proof of availability sends a random committee of peers to re-audit, and both a faker and a rubber-stamper get slashed. Metering signs every unit of work, so earnings are units times weight times reputation. On-chain uptime is committed to a public Merkle registry anyone can check against the chain.
Four checks, each catching a different kind of lie.

Four separate checks, each one designed to catch a different kind of lie. Proof of storage challenges a node on the actual bytes it claims to be holding, a math check rather than its word for it. Proof of availability sends a random committee of other nodes to re-audit, and both a faker and anyone who rubber-stamps a faker get penalized for it. Metering signs every unit of work, so what a node earns is units times weight times reputation and nothing else. Uptime is committed to a public registry that anyone can check against the chain.

A node cannot fake the work. Serve the wrong bytes, go dark, or lie about uptime and the proof fails, so it earns nothing. The coordinator hands out the jobs, but it cannot invent the results, the network proves them.

Serve the wrong bytes, go dark, or lie about being online, and the proof simply fails. A node that cheats does not earn a smaller amount that week. It earns nothing. That gap, between getting paid for provable work and getting listed for an address, is the entire difference between a real distributed network and a map of IP addresses somebody promised were busy.

So, is there a public GitHub?

That was the other half of the question, so, plainly: no, not for the platform itself. The code is private for now, and it will most likely open up further after the third-party audit. We are not going to dress that up or pretend a repo is public when it is not.

What we will commit to is this. On mainnet, every contract will be enumerated and available for inspection, so the parts that decide who gets paid and how the money moves will be readable on-chain by anyone who wants to check our arithmetic against what the network actually does. Distributed, for us, is a statement about where the work runs and how it gets verified. Whether every line of code is public is a separate question, and we would rather answer both of them straight than blur them into one comfortable word.

If you want the whole picture on a single page, the full architecture map is here. And if you have the next skeptical question, come join us in the Discord, it is the right place to put it. Asking how the machine really works is the correct instinct, and we would much rather earn that answer than ask you to take it on faith.

Share this
Join the Discord Read the white paper More news →