The views expressed in this article are those of the author and do not necessarily reflect the position of CoinGeek.
Last week, on August 10, 2026, I argued on CoinGeek that Bitcoin indexers are good: token state belongs with indexers and overlays, never in node queries. That argument was one ring of a bigger structure called the Mandala network, and the mandala scales because a break anywhere in the stack is a local event, never a global one. A dead indexer strands nobody’s consensus. A crashed node breaks nobody’s app. An application that rugs its own users never touches the chain at all.
That containment is separation of concerns doing its job, and it is the entire survival strategy of the design.
The BSV Skills Center documents the Mandala as exactly three layers: a Teranode consensus core, an overlay service layer, and an application layer. Three rings, three jobs, and no ring doing another ring’s work.
*Order where order pays, chaos where chaos is safe: a dense disciplined core, a deliberate middleware mesh, and a gloriously messy edge. Trace any single cut in that drawing and follow the damage. It never leaves the neighborhood.*
The core is solid. The doors are the work
Assume global Teranodes in play, and the consensus layer is really, really solid: BSV Association reported over 1 million transactions per second sustained across a two-week distributed testnet trial, and the first public Teranode release shipped on October 14, 2025.
A million per second, for two weeks!
The core’s shape has a name, too: Duncan Watts and Steven Strogatz described small-world networks in Nature in June 1998, short paths between any two points, dense clustering among neighbors. A tight cluster of well-connected heavyweight nodes is less a decentralization failure and more the small-world core doing exactly what the topology says it should.
The work is at the doors.
Ingestion (ARC and/or Arcade) needs more endpoints and faster P2P validation, and subtree and block announcement can still go a few directions. Teranode already propagates blocks as subtrees, batches of transaction IDs announced continuously while the block is still being built, so the open question is how those announcements travel at scale, and something like Cloudflare offers one answer while the rest of the edge-worker platforms offer variations on the theme…
At GorillaPool, we are building our own answer on IPv6 multicast tunnels, a project we are calling MonkeyCast, and it is in active development, which is another way of saying it does not fully exist yet. If it works, we will probably move Arcade and node communications into MonkeyCast tunnels as well, for scalability and efficiency. My sketch of the core at scale: 6 to 12 ingestion endpoints, never more than about 100 individual nodes, all directly connected by tunnels, with hyperscaler, commercial-grade connectivity.
A backbone, on purpose.
Back to the top ↑
Middleware in the middle, chaos at the edge
One ring out from the core is where overlays and indexers live. JungleBus is ours at GorillaPool: an overlay engine that indexes the whole chain so applications never have to. And as we established last week, bitcoin-based tokens need an indexer that MUST live somewhere in the stack, and it should absolutely not query nodes directly for state, because that model does not scale. State lives here, in ring two, held by distributed peers for persistence and for security.
Who runs ring two? Mining pools, almost exclusively, operate as what I started calling a “Web3 ISP” in 2022. The term is mine, and I am keeping it! An ISP does not care what you say on the call; it sells a reliable pipe with a service agreement and a support line attached. “Miners” in the BSV space need to adopt that posture, because app developers cannot build a business on a hobby endpoint that disappears when its operator gets bored. I expect ring two to settle out as dozens, maybe hundreds of middleware services, each with a name on the door and liability attached to the name.
The outer ring is wallets and applications, and in any bitcoin-based system, those are intrinsically connected pieces of software. Everything from healthcare hashing systems to video games lives out here. Each app holds only the data it needs, in a light index on its own servers, queries ring two when that beats holding more, and spends its resources on exactly one thing: its own customers. Millions of apps, eventually, and never one of them “bothering” the nodes at the consensus layer.
The edge can also skip the stack entirely.
In the true peer-to-peer SPV model, the sender hands the transaction directly to the recipient and broadcasts nothing. The recipient broadcasts optionally, on their own schedule, based entirely on their personal or company risk model. A payment becomes a conversation between two parties instead of an announcement to the whole planet.
If this shape looks familiar, it should. The internet scales because IP is a narrow, stable waist that nobody renegotiates, and innovation piles up above it (TCP, HTTP, the entire application universe) and below it (fiber, 5G) while the waist holds still. The mandala is the same drawing with different labels: a set-in-stone consensus waist, commercial plumbing above it, and a riot of applications above that. It is modeled on the internet stack because the internet stack is the only architecture in history that has actually scaled.
Back to the top ↑
Isn’t this just trusted third parties with extra steps?
If millions of apps query middleware run by mining pools, haven’t we just reinvented the trusted middleman and given him a cooler job title?
No, and the reason is SPV. An application holds block headers and checks Merkle proofs against them itself. Middleware serves the data; the math decides whether the data is true.
Proofs, not promises.
Does an app have to take GorillaPool’s word for anything? No. Does a lying indexer get caught? By the first client that checks a proof. Is one middleware provider a chokepoint? One would be, which is exactly why the design calls for dozens to hundreds of competitors with reputations and liability attached, racing each other on price and uptime.
Now flip the question, because the alternative is the actual centralizing force. If every app queries consensus nodes for state, only a handful of machines on Earth could ever carry that load, and everyone ends up standing in line at the same few windows. That is a mainframe with extra steps, and there’s a reason the internet got away from the model starting in the 1980’s and almost completely replaced it over twenty years ago.
So here is the ask, aimed squarely at BSV miners: adopt the Web3 ISP posture that GorillaPool has done.
WE NEED YOU TO DO THIS TOO!
Stand up the endpoints, run the overlays, publish the service agreements, and become the rock-solid, set-in-stone second ring that developers can build on without asking anyone’s permission. The core has shown its numbers. The edge shows up on its own the moment the middle is worth building on.
Cut the mandala anywhere, and nothing dies.
This opinion piece is published to encourage discussion. The author’s views are their own and do not constitute legal, procurement, or policy advice, nor do they represent the positions of CoinGeek or its partners.
Back to the top ↑
Watch: Want to develop on BSV? Here’s how you can build with Mandala
Be the first to comment