Summary
BuilderNet is Flashbots’ TEE-based shared block building system (Phase 1), operated jointly by Flashbots, Beaver Builder, and Nethermind. It represents the first step toward SUAVE (Phase 3) — a globally distributed, permissionless MEV marketplace. The decentralization roadmap uses four phases to measure progress away from centralized monolithic builders.
The Four Phases
| Phase | Name | Description | Test |
|---|---|---|---|
| 0 | Status Quo | Single company’s cloud; all blocks attributable to one entity | Can you attribute blocks to a specific endpoint? |
| 1 | Replicated Privacy | TEE-based sharing; crash fault tolerant across N parties | Identical block output regardless of which single party fails |
| 2 | Modular Distributed Building | BFT co-building; blocks can no longer be attributed to a single entity | Blocks co-built in distributed manner, proofs verifiable by any node |
| 3 | Global Parallel Building | Meaningful building activity worldwide; any machine can join/exit | Nodes in any region can economically influence block outputs |
Phase 3 = SUAVE: a Single Unified Auction for Value Expression — a globally distributed, permissionless MEV network.
Flashbots is transitioning Phase 0 → Phase 1. Phase 2 is in experimental/research stage.
BuilderNet Architecture (Phase 1)
Operators: Flashbots, Beaver Builder, Nethermind
Technology: Trusted Execution Environments (TDX/SGX)
Key properties:
- Multiple operators run identical builder code inside TEEs.
- TEEs allow attestation: external parties can verify what code ran and what data was input.
- Order flow can be shared with privacy guarantees — a TEE enforces that only valid operations (e.g., backruns) are performed on shared transaction data, preventing frontrunning.
- Crash fault tolerance: if one operator goes down, others produce the same block output.
Value Distribution / Refund Mechanism
After each block, a simulation calculates the marginal value contribution of each order flow source. The surplus above these contributions stays in the system; the rest is redistributed to order flow sources.
Open problem: how to reward infrastructure operators (not just OFA providers) in a programmatic, verifiable way. No satisfying mechanism exists as of April 2026.
SUAVE Clarification
SUAVE is the long-term vision; BuilderNet is one instantiation. Relationship is like “world computer” (Ethereum’s aspiration) vs. the beacon chain (current implementation).
Flashbots Product Stack
| Product | Purpose | Phase |
|---|---|---|
| BuilderNet | TEE-based shared building | 1 |
| Signal-Boost | Secure sandboxing; co-location for untrusted parties to build block segments | 2 |
| rblib | Open-source library of modular block building primitives | 2 |
| DCEA / Proof of Cloud | Bind TEEs to physical deployments; verifiable infrastructure | 2 |
| Mosaik | Self-organizing P2P topologies; metadata-based routing without central control | 2–3 |
| NAMP/ADCNet | Network Anonymized Mempools; anonymous broadcast (builds on ZIPNet) | 2–3 |
| Flashnet | Anonymous broadcast with IBLT-based auction scheduling | 3 |
Flashnet / ADCNet (Anonymous Broadcast)
Building on ZIPNet, Flashnet aims for anonymous broadcast with:
- Lower latency than existing protocols (Tor, etc.)
- Strong anonymity guarantees
- Bandwidth sufficient for blockchain use cases
Auction-based scheduling: uses Invertible Bloom Lookup Tables (IBLTs) to transmit message metadata (hash, size, utility) while preserving anonymity. Clients run a knapsack solver on aggregated metadata to determine who gets bandwidth. This maximizes total utility moving through the channel, unlike ZIPNet’s random scheduling which wastes bandwidth on collisions.
Why Decentralization Requires Privacy
The post argues privacy is the key to harmonizing:
- User needs: users can only harness their information by keeping it private from intermediaries (principal-agent problem).
- Distributed security: ordering manipulation can happen at any layer, even before transaction submission — only programmable privacy prevents this end-to-end.
- Scaling: true scaling under MEV contention requires modular, distributed builders.
Both cypherpunks and institutions need the same privacy properties — they are universal.
Geographic Decentralization (Phase 3)
The biggest open question: how do nodes outside the US/EU compete when latency is a direct competitive advantage?
Current BuilderNet: US, EU, Tokyo nodes. Phase 3 requires meaningful building activity worldwide — “far regions” shouldn’t just follow central regions’ output without being able to economically affect it.
Related Pages
- PBS and MEV-Boost — Current PBS architecture this is replacing
- ePBS: Enshrined Proposer-Builder Separation (EIP-7732) — Protocol-level changes that must be supported by decentralized infra
- TEEs and Attested TLS — TEE technology underpinning Phase 1
- MEV Supply Chain: Searchers, Builders, Relays, and Validators — Actor landscape
❓ Flashnet/NAMP/ADCNet (Phase 2–3 anonymous broadcast) is described in the body of this page. No dedicated wiki page exists yet; create one once more source material is available.
Key Sources
- Decentralized Building: wat do? (Flashbots, Feb 2026) — phases, roadmap, products, why privacy matters
- What Emerged from the Blockspace Forum Workshop in Cannes (Apr 2026) — BuilderNet TEE presentation; value distribution open problem
- Anonymous broadcast with auction-based scheduling (Flashbots, Mar 2026) — IBLT-based Flashnet scheduling design