Microsoft's Magnetar makes proofs 10 to 39 times faster, in research

A computer result usually arrives with a question attached: did the program really run as claimed? Today the answer rests on trust in whoever ran it, backed by contracts, audits and logs. Zero-knowledge proofs offer another route. The program runs, and alongside its output it produces a short cryptographic certificate. Anyone holding it can confirm the result in milliseconds, without rerunning the program and without seeing the data it used.

The idea dates from the 1980s, but three obstacles kept it from becoming a general tool. Producing a proof has historically taken millions of times more computation than running the program itself. Many systems also required developers to rewrite their program by hand as a cryptographic circuit, a fixed arithmetic description of one exact computation, a job that takes months and must be redone with every change. And many proofs rest on elliptic curves, which a sufficiently powerful quantum computer could break. Microsoft's Quantum Safe Program will move its products and services to post-quantum cryptography, a goal it recently accelerated to 2029.

Microsoft Research has worked on the problem for more than a decade. Its 2019 system Spartan returned to the sum-check approach and made it work for general programs without a trusted setup, a secret initial step whose leak would let anyone forge proofs. Phalanx, in 2021, exploited repetition, such as a processor running the same instruction cycle at every step, to prove computations an order of magnitude faster. Magnetar, announced on 15 October 2026, proves programs written in Rust, Go or C and rests its security only on standard hash functions.

The headline figure is that Magnetar runs 10 to 39 times faster on average than the prior fastest comparable systems. Those numbers come from Microsoft Research's own announcement. The post describes a research system, and it does not mention any product built on it.

For AI agents acting on someone's behalf, the idea is that each action could carry a proof that it complied with the policy it was given. For regulated financial institutions, it would show that transactions were processed correctly without making them public. Microsoft Research presents both uses as possible once producing a proof is as routine as running the program, and it does not put a date on that point.