Nescience State Separation Architecture (NSSA) is a programmable blockchain system that introduces a clean separation between public and private states, while keeping them fully interoperable. It lets developers build apps that can operate across both transparent and privacy-preserving accounts. Privacy is handled automatically by the protocol through zero-knowledge proofs (ZKPs). The result is a programmable blockchain where privacy comes built-in.
Typically, public blockchains maintain a fully transparent state, where the mapping from account IDs to account values is entirely visible. In NSSA, we introduce a parallel *private state*, a new layer of accounts that coexists with the public one. The public and private states can be viewed as a partition of the account ID space: accounts with public IDs are openly visible, while private accounts are accessible only to holders of the corresponding viewing keys. Consistency across both states is enforced through zero-knowledge proofs (ZKPs).
Public accounts are represented on-chain as a visible map from IDs to account states and are modified in-place when their values change. Private accounts, by contrast, are never stored in raw form on-chain. Each update creates a new commitment, which cryptographically binds the current value of the account while preserving privacy. Commitments of previous valid versions remain on-chain, but a nullifier set is maintained to mark old versions as spent, ensuring that only the most up-to-date version of each private account can be used in any execution.
Our goal is to enable full programmability within this hybrid model, matching the flexibility and composability of public blockchains. Developers write and deploy programs in NSSA just as they would on any other blockchain. Privacy, along with the ability to execute programs involving any combination of public and private accounts, is handled entirely at the protocol level and available out of the box for all programs. From the program’s perspective, all accounts are indistinguishable. This abstraction allows developers to focus purely on business logic, while the system transparently enforces privacy and consistency guarantees.
To the best of our knowledge, this approach is unique to Nescience. Other programmable blockchains with a focus on privacy typically adopt a developer-driven model for private execution, meaning that dApp logic must explicitly handle private inputs correctly. In contrast, Nescience handles privacy at the protocol level, so developers do not need to modify their programs—private and public accounts are treated uniformly, and privacy-preserving execution is available out of the box.
### Example: creating and transferring tokens across states
To achieve both state separation and full programmability, NSSA adopts a stateless program model. Programs do not hold internal state. Instead, all persistent data resides in accounts explicitly passed to the program during execution. This design enables fine-grained control over access and visibility while maintaining composability across public and private states.
### Execution types
Execution is divided into two fundamentally distinct types based on how they are processed: public execution, which is executed transparently on-chain, and private execution, which occurs off-chain. For private execution, the blockchain relies on ZKPs to verify the correctness of execution and ensure that all system invariants are preserved.
Both public and private executions of the same program are enforced to use the same Risc0 VM bytecode. For public transactions, programs are executed directly on-chain like any standard RISC-V VM execution, without generating or verifying proofs. For privacy-preserving transactions, users generate Risc0 ZKPs of correct execution, and validator nodes only verify these proofs rather than re-executing the program. This design ensures that from a validator’s perspective, public transactions are processed as quickly as any RISC-V–based VM, while verification of ZKPs keeps privacy-preserving transactions efficient as well. Additionally, the system naturally supports parallel execution similar to Solana, further increasing throughput. The main computational bottleneck for privacy-preserving transactions lies on the user side, in generating zk proofs.
### Resources
- [IFT Research call](https://forum.vac.dev/t/ift-research-call-september-10th-2025-updates-on-the-development-of-nescience/566)
- [NSSA Token program desing](https://www.notion.so/Token-program-design-2538f96fb65c80a1b4bdc4fd9dd162d7)
- [NSSA cross program calls](https://www.notion.so/NSSA-cross-program-calls-Tail-call-model-proposal-extended-version-2838f96fb65c8096b3a2d390444193b6)
> **Note for Fedora 41+ users:** GCC 14+ has stricter C++ standard library headers that cause build failures with the bundled RocksDB. You must set `CXXFLAGS="-include cstdint"` when running cargo commands. See the [Run tests](#run-tests) section for examples.
This tutorial walks you through creating accounts and executing NSSA programs in both public and private contexts.
> [!NOTE]
> The NSSA state is split into two separate but interconnected components: the public state and the private state.
> The public state is an on-chain, publicly visible record of accounts indexed by their Account IDs
> The private state mirrors this, but the actual account values are stored locally by each account owner. On-chain, only a hidden commitment to each private account state is recorded. This allows the chain to enforce freshness (i.e., prevent the reuse of stale private states) while preserving privacy and unlinkability across executions and private accounts.
>
> Every piece of state in NSSA is stored in an account (public or private). Accounts are either uninitialized or are owned by a program, and programs can only modify the accounts they own.
>
> In NSSA, accounts can only be modified through program execution. A program is the sole mechanism that can change an account’s value.
> Programs run publicly when all involved accounts are public, and privately when at least one private account participates.
> New accounts begin in an uninitialized state, meaning they are not yet owned by any program. A program may claim an uninitialized account; once claimed, the account becomes owned by that program.
> Owned accounts can only be modified through executions of the owning program. The only exception is native-token credits: any program may credit native tokens to any account.
> However, debiting native tokens from an account must always be performed by its owning program.
In this example, we will initialize the account for the Authenticated transfer program, which securely manages native token transfers by requiring authentication for debits.
Now that we have a public account initialized by the authenticated transfer program, we need to fund it. For that, the testnet provides the Piñata program.
> The new account is uninitialized. The authenticated transfers program will claim any uninitialized account used in a transfer. So we don't need to manually initialize the recipient account.
> Private accounts are structurally identical to public accounts; they differ only in how their state is stored off-chain and represented on-chain.
> The raw values of a private account are never stored on-chain. Instead, the chain only holds a 32-byte commitment (a hash-like binding to the actual values). Transactions include encrypted versions of the private values so that users can recover them from the blockchain. The decryption keys are known only to the user and are never shared.
> Private accounts are not managed through the usual signature mechanism used for public accounts. Instead, each private account is associated with two keypairs:
> Just like public accounts, private accounts can only be initialized once. Any user can initialize them without knowing the owner's secret keys. However, modifying an initialized private account through an off-chain program execution requires knowledge of the owner’s secret keys.
> Transactions that modify the values of a private account include a commitment to the new values, which will be added to the on-chain commitment set. They also include a nullifier that marks the previous version as old.
> The nullifier is constructed so that it cannot be linked to any prior commitment, ensuring that updates to the same private account cannot be correlated.
For now, focus only on the account id. Ignore the `npk` and `ipk` values. These are the Nullifier public key and the Viewing public key. They are stored locally in the wallet and are used internally to build privacy-preserving transactions.
Also, the account id for private accounts is derived from the `npk` value. But we won't need them now.
This command will run the Authenticated-Transfer program locally, generate a proof, and submit it to the sequencer. Depending on your machine, this can take from 30 seconds to 4 minutes.
As a general rule, private accounts can only be modified through a program execution performed by their owner. That is, the person who holds the private key for that account. There is one exception: an uninitialized private account may be initialized by any user, without requiring the private key. After initialization, only the owner can modify it.
This mechanism enables a common use case: transferring funds from any account (public or private) to a private account owned by someone else. For such transfers, the recipient’s private account must be uninitialized.
#### Sending tokens from the public account to a private account owned by someone else
For this tutorial, we’ll simulate that scenario by creating a new private account that we own, but we’ll treat it as if it belonged to someone else.
Let's create a new (uninitialized) private account like before:
Now we'll ignore the private account ID and focus on the `npk` and `ipk` values. We'll need this to send tokens to a foreign private account. Syntax is very similar.
The command above produces a privacy-preserving transaction, which may take a few minutes to complete. The updated values of the private account are encrypted and included in the transaction.
Once the transaction is accepted, the recipient must run `wallet account sync-private`. This command scans the chain for encrypted values that belong to their private accounts and updates the local versions accordingly.
#### Transfers in other combinations of public and private accounts
We’ve shown how to use the authenticated-transfers program for transfers between two public accounts, and for transfers from a public sender to a private recipient. Sending tokens from a private account (whether to a public account or to another private account) works in essentially the same way.
### The token program
So far, we’ve made transfers using the authenticated-transfers program, which handles native token transfers. The Token program, on the other hand, is used for creating and managing custom tokens.
> The token program is a single program responsible for creating and managing all tokens. There is no need to deploy new programs to introduce new tokens. All token-related operations are performed by invoking the appropriate functions of the token program.
> The Token program manages its accounts in two categories. Meaning, all accounts owned by the Token program fall into one of these types.
> - Token definition accounts: these accounts store metadata about a token, such as its name, total supply, and other identifying properties. They act as the token’s unique identifier.
> - Token holding accounts: these accounts hold actual token balances. In addition to the balance, they also record which token definition they belong to.
Like any other private account owned by us, it cannot be seen by other users.
#### Custom token transfers
The Token program has a function to move funds from one token holding account to another one. If executed with an uninitialized account as the recipient, this will be automatically claimed by the token program.
The transfer function can be executed with the `wallet token send` command.
Let's create a new public account for the recipient.
NSSA includes an AMM program that manages liquidity pools and enables swaps between custom tokens. To test this functionality, we first need to create a liquidity pool.
#### Creating a liquidity pool for a token pair
We start by creating a new pool for the tokens previously created. In return for providing liquidity, we will receive liquidity provider (LP) tokens, which represent our share of the pool and are required to withdraw liquidity later.
To hold these LP tokens, we first create a new account:
```bash
wallet account new public
# Output:
Generated new account with account_id Public/FHgLW9jW4HXMV6egLWbwpTqVAGiCHw2vkg71KYSuimVf
```
Next, we initialize the liquidity pool by depositing tokens A and B and specifying the account that will receive the LP tokens:
If you inspect the `user-holding-a` and `user-holding-b` accounts passed to the `wallet amm new` command, you will see that 100 and 200 tokens were deducted, respectively. These tokens now reside in the liquidity pool and are available for swaps by any user.
#### Swaping
Token swaps can be performed using the wallet amm swap command:
Once executed, 5 tokens are deducted from the Token A holding account and the corresponding amount (determined by the pool’s pricing function) is credited to the Token B holding account.
#### Withdrawing liquidity from the pool
Liquidity providers can withdraw assets from the pool by redeeming (burning) LP tokens. The amount of tokens received is proportional to the share of LP tokens being redeemed relative to the total LP supply.
This operation is performed using the `wallet amm remove-liquidity` command:
This instruction burns `balance-lp` LP tokens from the user’s LP holding account. In exchange, the AMM transfers tokens A and B from the pool’s vault accounts to the user’s holding accounts, according to the current pool reserves.
The `min-amount-a` and `min-amount-b` parameters specify the minimum acceptable amounts of tokens A and B to be received. If the computed outputs fall below either threshold, the instruction fails, protecting the user against unfavorable pool state changes.
#### Adding liquidity to the pool
Additional liquidity can be added to an existing pool by depositing tokens A and B in the ratio implied by the current pool reserves. In return, new LP tokens are minted to represent the user’s proportional share of the pool.
This is done using the `wallet amm add-liquidity` command:
In this instruction, `max-amount-a` and `max-amount-b` define upper bounds on the number of tokens A and B that may be withdrawn from the user’s accounts. The AMM computes the actual required amounts based on the pool’s reserve ratio.
The `min-amount-lp` parameter specifies the minimum number of LP tokens that must be minted for the transaction to succeed. If the resulting LP token amount is below this threshold, the instruction fails.
