# Harmonic Brings Aristotle Formal Verification to C and Verilog

Published 2026-10-28 · AI.info Team · AI.info

> Harmonic has expanded its Aristotle AI system to verify C software and Verilog hardware designs. The company says the service is available with consumption-based pricing, while its terms limit proof claims to the specifications and assumptions used.

[Harmonic](https://www.harmonic.fun/news/aristotle-software-hardware-verification/) says its Aristotle system can prove software and hardware properties across all permitted inputs; its terms say a “verified” result applies only to the specification, model and assumptions used, and still requires independent review. The company announced the expansion on October 27, adding support for C and Verilog after building Aristotle for mathematical reasoning.

The distinction matters for a product aimed at checking code and chip designs: a proof can establish that a property holds within a defined system, but it does not certify that the specification captures every real-world requirement. Harmonic is pitching Aristotle as a way to make formal methods easier to use, while its own terms place responsibility for validating results on customers.

## Aristotle proposes properties, then reports what it can prove

Aristotle proposes properties from submitted code and reports which ones it proves, which it refutes and what assumptions each result depends on. Engineers review those properties, commit them to a repository and can run the system in continuous integration on each commit.

For C, the company describes checks against specified requirements, including functional correctness and the absence of certain classes of undefined behavior. For Verilog, it says Aristotle handles control-path properties such as liveness and safety, alongside mathematical properties of data paths. The announcement does not publish general benchmark scores or a comparison with established formal-verification tools.

## Tensordyne test reaches 110,000 cycles

Harmonic’s Verilog case study centers on a Tensordyne design block it describes as particularly difficult to verify. The model contained 12.2 million gate equivalents, the company says, and Aristotle assessed 241 whole-block properties: it proved 185 and refuted 56, each with a counterexample.

The reported bugs appeared 25,000 to 110,000 cycles after reset, beyond a memory-initialization period of about 24,600 cycles. The work helped Tensordyne find eight bugs early in the design cycle, Harmonic says. Those figures come from Harmonic’s account of the project; the company’s page also says the proofs cover the modeled partition and the input sequences allowed by its environment.

## LibTomMath speedup has a defined proof boundary

On the C side, Harmonic says Aristotle repaired three bugs in LibTomMath version 1.2.0 and optimized two arithmetic routines. The company reports modular exponentiation running 31% to 38% faster in its benchmarks, with the optimized routines proved against the same specifications as the repaired code.

The case study narrows what that result means: selected routines and helpers were proved for stated behaviors and assumptions, while full modular exponentiation was benchmarked, not proved. The tests used an Intel Xeon Platinum 8375C with GCC 14.2, and performance claims apply to the measured workloads and configuration, Harmonic says.

## Terms keep human review in the loop

Harmonic CEO Tudor Achim framed the product as a way to catch defects before they grow costly.

> “For businesses that rely on software and hardware, mathematically verifying critical systems provides a new way to identify defects and vulnerabilities before they become costly or catastrophic problems.”
>
> [Tudor Achim, co-founder and CEO of Harmonic](https://www.harmonic.fun/news/aristotle-software-hardware-verification/)

The company’s terms add a firm qualification: Harmonic does not guarantee the validity or accuracy of its output, and customers must independently review, test and validate it. For high-risk applications, the terms require qualified human review and say Aristotle’s output cannot serve as the sole basis for design, safety, certification or deployment decisions.

Aristotle is available through self-serve, consumption-based pricing, Harmonic says. Enterprise plans include volume pricing, dedicated support and on-premises or private-cloud deployment; the announcement does not state a per-unit rate.

## Sources

1. Original source [Harmonic · Aristotle software hardware verification](https://www.harmonic.fun/news/aristotle-software-hardware-verification/) harmonic.fun

Other pages read

2. [Harmonic · page 2](https://www.harmonic.fun/aristotle/case-studies/tensordyne/) harmonic.fun
3. [Harmonic · page 3](https://www.harmonic.fun/aristotle/case-studies/libtommath/) harmonic.fun
4. [Harmonic · page 4](https://www.harmonic.fun/legal/terms/) harmonic.fun

## Related stories

- [Axiom Formalizes the 15,000-Page Classification of Finite Simple Groups](https://ai.info/news/axiom-formalizes-classification-finite-simple-groups-lean)
- [Anthropic Uses Claude to Prove 239 Tests Across 123 Libraries](https://ai.info/news/anthropic-claude-formal-verification-123-libraries)
- [Microsoft Research Used PDD Kit to Formally Verify Login Logic in a Rust Web App](https://ai.info/news/microsoft-research-pdd-kit-rust-web-app)
- [MAGS Verifies Agent-Written Code, Then Exposes Its Limits](https://ai.info/news/mags-verifies-agent-written-code-then-exposes-its-limits)
- [Chrome’s Gmail-Sourced Autofill Requires Personal Intelligence](https://ai.info/news/chrome-gmail-ai-autofill-suggestions)

## Explore

[More articles](https://ai.info/news)

---

Canonical: https://ai.info/news/harmonic-aristotle-formal-verification-c-verilog
