Litepaper Download PDF

MeterCall L4 Whitepaper

Version 1.0 · April 21, 2026 · Author: Yoshi (pseudonymous) · Network: Base (Layer 2), L4 Aggregation Layer · Token: PCP — 1,000,000,000 fixed supply, zero inflation

Abstract

MeterCall L4 is an aggregation and routing layer that sits above existing L1s, L2s, and L3s. Its purpose is narrow and stubborn: let anyone, anywhere, call any of 20 million+ Web2 APIs and 30+ blockchains through a single call, and let the person closest to the work get paid for routing it. The little guy with a Raspberry Pi in a bedroom beats the data center because he is closer, cheaper, and not trying to win everything. This paper explains how the routing works, how we keep the rich from capturing it, how PCP aligns the whole thing, and what we still don't know.

1. The Problem I Lived

I ran nodes. I ran a lot of nodes. And the people with data centers won every time.

It didn't matter that I showed up early. It didn't matter that I cared about the protocol. Once the economics turned, the people with capital bought bigger machines, put them in colocation, bought fatter pipes, and ate the rewards. The protocol said "decentralized." The reality was a small number of very well-capitalized operators, clustered in three or four ASNs, running most of the network.

That is not decentralization. That is a shell with rich-guy insides.

The pattern repeats on every chain I've touched:

This paper is about building a network where Joe Schmoe with a Raspberry Pi in Tulsa actually beats the data center in Ashburn — because he's closer to the request and the protocol rewards him for it.

2. Why Previous L1s, L2s, and L3s Didn't Fix It

L1s (Ethereum, Solana, Cosmos): honest attempt at base-layer decentralization. Validator sets still concentrate. Fees and MEV concentrate harder. None were designed for calling Web2.

L2s (rollups): cheaper execution. Still need sequencers. Sequencers are centralized today by design, decentralized "later." Still don't touch Web2.

L3s (app-chains): cheaper app-specific execution. Great for games. Doesn't solve "I want to call the Stripe API and pay with crypto from a chain that isn't Ethereum."

None of these layers compete on where the request is served from. They compete on block production and execution. The actual bottleneck for real-world usage — the 50ms to 400ms of network latency between a user and the data that answers their question — is treated as somebody else's problem.

MeterCall L4 treats that latency as the whole problem.

3. Architecture

3.1 What L4 actually is

L4 is an aggregation and routing overlay. It does not produce blocks. It does not settle state. It sits on top of Base (L2 Ethereum) for settlement, receipts, and dispute resolution, and it routes calls between any caller and any of three pools of endpoints:

  1. Web2 APIs — 20M+ indexed endpoints. REST, GraphQL, gRPC, WebSocket.
  2. Blockchains — 30+ chains via native RPC + a bridge signer mesh.
  3. Builder modules — programs deployed by creators that compose the first two.

Every call is priced, metered, and billed per-invocation. x402 for payment-required responses. EIP-712 for signed routing receipts.

3.2 Geo-routing: why Joe Schmoe wins

For a call from client c to endpoint e, the scheduler picks the node n that minimizes expected total cost:

cost(n, c, e) =  α · rtt(n, c)
              +  β · rtt(n, e)
              +  γ · price(n)
              -  δ · reputation(n)
              -  ε · residential_bonus(n)
              +  ζ · concentration_penalty(n)

Default coefficients (governance-tunable): α=1.0, β=1.0, γ=0.6, δ=0.8, ε=0.4, ζ=1.2. rtt is round-trip-time in ms. price is the node's posted PCP-per-call. reputation is a rolling success-weighted score. residential_bonus is +20% scheduling weight for non-datacenter ASNs. concentration_penalty scales quadratically with how many other active nodes share the node's ASN and geohash-5 cell.

A residential node in Tulsa serving a Tulsa caller, at the same price as an Ashburn datacenter node serving the same Tulsa caller, wins on α, ε, and ζ. It is not a handicap. It is the protocol acknowledging that proximity is real work.

                  scheduler decision (simplified)
                  -----------------------------------

   caller(Tulsa) ----> [ candidate set ]
                         |
                         +-- node A: Tulsa residential, ASN 7018,  price 1.00
                         +-- node B: Ashburn datacenter, ASN 16509, price 0.90
                         +-- node C: Dallas datacenter,  ASN 16509, price 0.95
                         +-- node D: Tulsa residential,  ASN 22773, price 1.05
                         |
                         v
                   score(A) = 12      <-- lowest, wins
                   score(B) = 18
                   score(C) = 17
                   score(D) = 13

3.3 Gossip layer

Nodes gossip three things and nothing else: (node_id, advertised_endpoints, price), (call_id, receipt_hash), and (node_id, heartbeat). Gossip is epidemic with fanout 6, TTL 4 hops. Anything larger than a receipt hash does not belong in gossip — it belongs in a direct fetch or on-chain.

       gossip fanout (one epoch, fanout=6, TTL=4)
       -------------------------------------------

       A ---> B C D E F G
                |
                B ---> H I J K L M
                         |
                         H ---> N O P Q R S ...   (TTL=1, stops here)

3.4 Anti-oligarch dilution curve

The thing that kills every "decentralized" network is: one entity accumulates nodes until they dominate. We fight this at two layers — hard and soft.

Hard cap. No single operator (proven by stake signature) can earn from more than 50 nodes in a 1,000-node network window. Excess nodes stay online but earn zero.

Soft curve. Effective reward weight per node decays as a function of ASN and geohash-5 concentration:

weight(n) = base(n) · (1 / (1 + k · asn_share(n)^2))
                    · (1 / (1 + k · geo_share(n)^2))

with k = 4 by default. A node in an ASN that already holds 20% of active nodes has its weight multiplied by 1 / (1 + 4·0.04) = 0.862. At 50% ASN share, 0.5. At 80% ASN share, 0.28.

       dilution curve (ASN share -> weight multiplier, k=4)
       ----------------------------------------------------

       1.00 |*****..
       0.90 |      ****.
       0.80 |          ***.
       0.70 |             **.
       0.60 |               **.
       0.50 |                 **.
       0.40 |                    **.
       0.30 |                       ***.
       0.20 |                           ****...
       0.10 |                                  *********
       0.00 +------------------------------------------>
            0%    20%    40%    60%    80%    100%  ASN share

The curve is continuous, so there's no cliff to game. A small operator with a unique ASN always beats a big operator adding their Nth redundant box.

3.5 Bridge mesh

Cross-chain calls route through a bridge-signer mesh. Signers are a rotating committee of the top-reputation nodes, weighted by dilution curve (so the committee cannot ossify into one ASN). Bridge receipts are EIP-712 signed, posted to Base for dispute, and optimistically finalized after a 20-minute challenge window.

4. PCP Tokenomics

PCP is the unit of account, the fee token, and the governance token. Fixed supply, zero inflation, deflationary by use.

4.1 Supply

4.2 Allocation

Bucket%TokensTerms
Team10%100,000,0004-yr vest, 1-yr cliff
Community / Airdrop30%300,000,000Launch + ongoing seasons
Treasury20%200,000,000DAO-controlled multisig
Launch liquidity20%200,000,000Paired on Base, locked 2 yr
Ecosystem grants15%150,000,000Milestone-based, DAO-voted
Strategic5%50,000,0002-yr vest, 6-mo cliff
TOTAL100%1,000,000,000

4.3 Burn mechanics

Every call settles in PCP (or in another asset converted to PCP at the settlement oracle). 30% of the PCP-denominated fee is burned. 70% is split:

If no builder module is invoked, the creator 25% rolls to the node.

4.4 Staking curve

Nodes must post S PCP as stake to be eligible for routing. Stake is slashable for provable misbehavior (failed-receipt disputes, signed conflicting bridge attestations, downtime beyond SLA). Routing weight scales as:

routing_weight(n) = dilution_weight(n) · sqrt(stake(n) / S_min)

sqrt not linear — a whale posting 100x stake gets 10x weight, not 100x. Combined with the dilution curve, this makes accumulation genuinely expensive.

5. Economic Security

5.1 Cost of attack

Assume an attacker wants to corner 51% of routing weight.

Total stake required: 500 nodes × 10,000 PCP × 2 (dilution offset) × sqrt-curve premium ≈ ~25M PCP locked, plus the operational cost of 500 nodes across 10+ ASNs in 10+ geohash-5 cells. At a conservative $0.10/PCP launch price, ~$2.5M in slashable stake plus infra. Corner case, not a target.

5.2 Sybil resistance

6. Roadmap

Q2 2026 (now — June):

Q3 2026 (July — September):

Q4 2026 (October — December):

7. Risks and Open Questions

I'm not going to pretend these are solved.

  1. Residential-ASN verification is imperfect. A determined adversary with a VPS fleet behind residential proxies can fake it for a while. Our answer: active probing from diverse vantage points, plus the economic cost of maintaining the fake. Not a proof. A cost curve.
  2. The dilution curve is a parameter, not a theorem. k=4 is a guess calibrated against simulated attacks. Governance can raise or lower it. It will probably need to move.
  3. Bridge-signer committees are still the weakest link in all of crypto. Rotation + dilution + slashing help, but a well-funded attacker coordinating half the committee remains a tail risk. We cap single-bridge TVL in the contract until we have more data.
  4. L2 dependency. Base could change fee markets, censor, or suffer outages. We target multi-L2 settlement (Base + Optimism + Arbitrum) in Q4, not today.
  5. Regulatory. PCP is a utility token within a functional network. That does not make it immune. The team is doxxed to counsel, not to the public.
  6. Scheduler gaming. If nodes learn to predict the scheduler, they can game reputation. We rotate coefficients seasonally and keep randomness in tie-breaking.
  7. Price volatility at TGE. 20% launch liquidity is aggressive but not infinite. Expect volatility. PCP is a utility token. Do not buy it for price.

8. How to Participate

Run a node. Pull the container. Post 10,000 PCP stake. Declare your endpoints. Start earning per-call. Residential +20%. See /run-a-node.html.

Stake. Delegate to nodes without running hardware. Stake-weighted routing means your pick earns more if you pick well. Slashing applies proportionally.

Build. Deploy a module. Creator earns 30% of every call to it. See /agent-build.html.

Govern. Hold PCP, vote on coefficients, burn rate, grant allocations, bridge caps. Treasury is 200M PCP under multisig that the DAO controls.

This is not a chain. This is a layer that makes every chain, and every API, reachable from one call, served by whoever is closest, owned by nobody who can big-time you. I lost to data centers once. I'm not losing again. Neither are you. — Yoshi, April 21, 2026