Harmonic's Aristotle can now formally verify Rust, and Solana SOL$74.41+1.6% programs are one of its first targets. Harmonic announced Rust support on 16 November 2026. Rust is the language most Solana programs are written in, and it joins the C and Verilog that Aristotle already covered.

Before launch, Aristotle produced 28 proved results on Manifest's open-source order book and turned up two issues, both since fixed. Two days later, a Solana Foundation team member said the Foundation is offering credits to teams that want critical code verified.

What Harmonic's Aristotle checks in Rust code

Formal verification uses mathematics to prove that a program does what its specification says, for every possible input. Testing checks the inputs someone thought to try. A proof covers all of them.

Harmonic says Aristotle can now prove that Rust code has "no undefined behavior and does exactly what its spec says, even if you use unsafe." Unsafe is the part of Rust where the compiler's usual safety checks are switched off. For teams porting C code to Rust, Aristotle can also prove the new code behaves the same as the original.

Harmonic is an AI mathematics company. Its Aristotle engine writes proofs in Lean, a proof language whose checker confirms each step. It is unrelated to the Solana block-building product of the same name. Robinhood chief executive Vlad Tenev is Harmonic's executive chairman and Tudor Achim its chief executive, per SiliconANGLE's report on its November 2025 funding round.

The Aristotle product page describes consumption-based pricing with no seats or licences. Harmonic said in its launch thread that it can run Aristotle on private cloud or on-premises for teams that cannot send code out, and it is giving new users their first $500 in credits.

What Aristotle proved about the Manifest order book on Solana

Manifest is a spot central limit order book on Solana: an exchange program that matches buy and sell orders by price and holds traders' deposits while it does. According to Harmonic's Manifest case study, Aristotle worked on Manifest's actual Rust code with no source changes. That included the unsafe red-black tree, a data structure that keeps orders sorted, at the core of the order book.

Aristotle produced 28 proved results across eight areas:

  • trading progress
  • error classification
  • order-book integrity
  • custody and recovery of funds
  • who can cancel an order
  • matching rules
  • reverse orders
  • quotes and rounding

The work found two issues. A quote function could let a fill overshoot its target at certain price ratios, after which the code tried to subtract two from one and hit an arithmetic error. The fix stops as soon as the fill meets its target.

Cancelling a global order also ran an unused quote conversion that could fail before the order was removed, and that path was deleted. Harmonic's own reply in its launch thread counted the result as "28 proofs, one real bug fixed", treating the second as a robustness fix.

The case study also sets the scope. The proofs assume the token and system programs behave as expected, and they do not bound compute units, the per-transaction processing budget on Solana.

Manifest's lead developer, unnamed on the page, said the trading-progress and error-classification proofs "go beyond the normal proof of no loss of funds and show un-brickable," adding: "That's what impressed me most." A bricked program is one stuck in a state where users can no longer trade or withdraw.

Manifest framed it the same way on X, saying the program is now "formally verified twice over": the first proof covered the safety of funds, and this one shows no one can stop users trading. Manifest's earlier verification was Certora's formal verification and manual audit of Manifest, dated 4 December 2024.

Solana Foundation credits for formal verification

The Solana Foundation is offering credits for formal verification of critical code, according to Foundation team member ludo. On 18 November 2026 he wrote that Harmonic worked closely with 15 ecosystem teams to calibrate Aristotle on the properties expected of an on-chain Solana program. He called Solana programs "a FIRST-CLASS citizen" in the release.

In a reply, he invited teams with critical code to get in touch: "@SolanaFndn is offering credits." The thread gives no amount or eligibility terms, so teams apply by contacting him directly.

He added that supporting "a healthy FV ecosystem" is one of his top priorities at the Foundation. The Foundation has started partnering with several providers that cover different parts of the stack, he wrote.

Certora, Runtime Verification and OtterSec: Solana's existing verification tools

Aristotle joins tools Solana teams already use. Certora's Prover checks programs against formal rules written by Certora's own team, an approach it laid out in its Breakpoint 2025 security talk. Runtime Verification built Kompass, which it used to check that the SPL Token program's Pinocchio rewrite stays safe. In 2023, OtterSec verified the Squads multisig program with the Kani model checker.

Harmonic's pitch is automation. Aristotle drafts the property statements and the proofs itself, working on unmodified Rust, and iterates until the Lean checker accepts them; the team then gets the statements to review. The Manifest case study shows that approach on a live Solana program holding real deposits.

Limits of AI formal verification on Solana programs

A proof is only as good as its specification and its model of the Solana runtime. OtterSec's 2023 write-up put it plainly: verification "won't stop you from specifying the wrong invariants or forgetting things," and cannot catch bugs outside the model, such as in the runtime.

That limit came up on launch day. A developer posting as alpharush replied that Aristotle's Solana model "may not be faithful" to how account data length is tracked, after the agent proved something it should have refuted.

The PetarMax account answered in the thread that the theorems' plain-English wording separates the serialised length field from the account data slice, and that the underlying proofs confirm it. When alpharush asked to see the formal theorems and their edit history, the same account replied that "we are working on all of that."

For Solana teams, three things are worth watching: the terms of the Foundation's credit offer, which of the 15 teams publish results, and whether Harmonic shares the Lean theorems behind its Solana case studies.

Keep up to date with the Solana eco
Follow us on Google News