Demand and allocation decisions
A market-entry track for pre-launch demand, price sensitivity, and allocation decisions without exposing raw responses.
PET Business Vision
One PET engine expanding into three market tracks
waLLLnut expands one FHE16-based PET engine into B2C demand intelligence, B2B risk intelligence, and blockchain confidential infrastructure.
PET Engine
Keep raw data hidden; expose only decision-ready results
waLLLnut combines FHE16, MPC, and threshold disclosure so sensitive data can be computed without moving raw records.
Identity and anomalous transaction risk
PET-based risk intelligence for cross-checking insurance, identity, account, and transaction patterns without moving source data.
Verifiable confidential infrastructure
The long-term expansion track connecting encrypted state, public verification, and selective disclosure to blockchain execution.
PET Landscape
From TEE to FHE, the PET landscape in plain terms
TEE is easy to start with, but it requires trust in hardware and operators. Cryptographic PET is harder to implement, but separates raw data and computation more strongly.
TEElow barrier, hardware trust
Raw data enters an enclaveFamiliar code runs insideOnly the result exits
The barrier is low, so many teams can build a PoC quickly. Differentiation is limited, and trust remains with the hardware vendor and operating environment.
Cryptographic PETprotocol and math-based trust
Raw data stays with each ownerThe protocol computes or provesOnly permitted output is revealed
MPC, PSI, PIR, ZK, FHE, and iO belong here. iO and FHE are especially hard to implement; iO remains outside practical product use, while optimized FHE can move into practical ranges.
Joint computation
Multiple parties compute one result while hiding each party's input.
Private set intersection
Find only overlapping customers, accounts, or events without revealing the rest.
Private retrieval
Let a user retrieve data without the server learning which item was requested.
Proof without disclosure
Prove a condition is satisfied without revealing the underlying secret.
Compute while encrypted
Difficult to implement, but integer arithmetic, bootstrapping, SIMD, and threading optimizations can bring it toward practical latency.
Hide the program itself
Conceptually powerful, but not practical yet for product workloads.
Easy to startCryptographic differentiation