RFC: Stake and Claiming Mechanism, Base Fees and Slashing

[en-tropyc]

Request for Comments (RFC) outlining the proposed economic mechanisms for staking requirements, fee structures, and slashing penalties. All feedback and suggestions are welcome.

Stake and Claiming Mechanism

Provers stake $BON to participate proving in the network. The amount of $BON they stake proportionally determines which executuion requests they have access to, as determined by the base fee of the execution request. We call this proportional access the "stake threshold". For example, with a stake threshold of 10x, a prover who stakes 100 $BON tokens can only complete workloads that cost 10 $BON or less.

Since in phase 1, since base fees will be paid in $USDC or $SOL, the stake threshold will be determined by the price of $BON in $USDC or $SOL. For example, with a stake threshold of 10x, a prover who stakes 100 $USDC or $SOL worth of $BON tokens can only complete workloads that cost 10 $USDC or 10 $SOL or less.

Here are some illustrative examples with a stake threshold of 10x:

Base Fee	Stake	Ratio (Stake/Base Fee)	Claimable?
$1	$10	10x	Yes
$1	$100	100x	Yes
$100	$10	0.1x	No
$100	$100	1x	No
$10	$101	10.01x	Yes

Open questions:

• How do we determine the stake threshold?
• How do we determine the base fee?
• Cap on staking? avoid centralization of revenue to 1 prover

Base Fees

Base fee is the minimum payment to a prover for completing a workload. In phase 1 base fees will be paid in $USDC or $SOL and in phase 2 base fees will be paid in $BON.

Base fees are determined by the cost of the workload. The base fee $F_b$ can be calculated by:

$$F_b = f(ZKP_s, C_{zkp})$$

where $ZKP_s$ is the size of the program to be proven and $C_{zkp}$ is the cost of computing 100,000 cycles of any a program. To start, this function will be a simple linear function:

$$f(ZKP_s, C_{zkp}) = \frac{ZKP_s}{100000} \times C_{zkp}$$

For example, if we have a program with:

• $ZKP_s = 500,000$ cycles
• $C_{zkp}$ = $0.1$ per 100,000 cycles

Then:
$$f(500000, 0.1) = \frac{500000}{100000} \times 0.1 = 50 \times 0.1 = 0.5$$

Note: Cost of computing cycles should be benchmarked on the minimum spec hardware of the prover network

Dynamic Base Fees

In the future, we may need to change the way the base fee is calculated. For example, under higher demand, we may want to increase the base fee to attract more demand. In such a case, we can formulate a new equation for the base fee as a polynomial function:

$$f(ZKP_s, C_{zkp}) = a_1(\frac{ZKP_s}{100000} \times C_{zkp}) + a_2(\frac{ZKP_s}{100000})^2 \times C_{zkp}^2 + a_3(\frac{ZKP_s}{100000})^3 \times C_{zkp}^3$$

Example Scenarios:

Demand Level	a₁	a₂	a₃	Description
Low	1.0	0.001	0.0001	Keeps fees close to the original linear model to attract more requests
Normal	1.0	0.01	0.001	Moderate scaling for complex computations
High	1.0	0.1	0.01	Significantly higher fees for complex computations
Extreme	1.0	0.5	0.1	Very aggressive scaling for complex computations

Sample calculation for ZKP_s = 500000, C_{zkp} = 0.1:

Demand Level	Calculation	Total
Low	0.5 + 0.00125 + 0.000125	0.50
Normal	0.5 + 0.0125 + 0.00125	0.51
High	0.5 + 0.125 + 0.0125	0.64
Extreme	0.5 + 0.625 + 0.125	1.25

Dynamic Base Fees: Coefficient Adjustment

The coefficients $(a_1, a_2, a_3)$ could be automatically adjusted based on network metrics:

1. Network Utilization Rate $(U)$

   • $U = \frac{\text{active provers}}{\text{total provers}}$
   • Or: $U = \frac{\text{current requests}}{\text{max throughput}}$
   • Example: If $U > 80$%, increase coefficients
   • If $U < 40$%, decrease coefficients

2. Queue Length $(Q)$

   • $Q = \frac{\text{pending requests}}{\text{average processing rate}}$
   • Longer queues → higher coefficients
   • Short/no queues → lower coefficients

3. Moving Average Formula
   For each coefficient $a_i$:
   $$a_i^{new} = a_i^{current} \times (1 + \alpha \times (U - U_{target}))$$
   where:

   • $\alpha$ is a dampening factor (e.g., 0.1)
   • $U_{target}$ is desired network load (e.g., 70%)

Slashing

When a prover is selected to complete a workload, they are required to complete the workload within a certain amount of time. If they do not complete the workload within the required time, they lose a portion of their stake (i.e. they are slashed).

• How much is slashed?
• How long is the timeout?