Terms of use — Numerical Research Kit
Publisher: Adem Vessell · Document version: 23 October 2026
These terms describe Numerical Research Kit, an open-source numerical comparison tool. They apply to the kit only, not to other Arkhē projects or services. The software is offered under the MIT License; this page does not add restrictions to its permissions.
What the kit provides
The kit compares baseline and candidate predictions against reference values you supply, summarizes mean squared error and declared computational costs, checks fixed-correction bounds, and creates and verifies saved comparison records.
You supply the references, cost figures, inputs and acceptance rules. The kit does not independently establish that these are correct, measure runtime or FLOPs, provide statistical confidence, or certify scientific conclusions. A passing record means the recorded comparison meets its specified checks; it is not a guarantee that a method is correct, faster or suitable for your application. Use appropriate domain-specific tests alongside it.
License and charges
The MIT License permits use, copying, modification, distribution and sale, subject to retaining the required copyright and permission notice. Its full text governs the software permissions and warranty disclaimer.
The kit itself requires no paid account, API key or payment to the publisher. Your execution environment or hosting application may have its own charges and terms. Those are separate from this kit.
Your inputs and results
Choose inputs you are entitled to use and share, and check that the numerical metric and acceptance rules suit the question. Review outputs before relying on them. Keep original evidence when changing inputs, software versions or acceptance rules; a new comparison should produce a new record.
The publisher claims no ownership of the data you provide or the results you generate. Software and documentation distributed with the kit remain subject to their license.
Privacy and execution environment
Version 0.3.1 of the package makes no network requests and contains no telemetry, analytics, accounts or hosted storage. Its maintainer receives no data through the package runtime. Output files contain supplied values and metadata alongside calculated results; they remain in the execution environment until you delete them.
Codex, ChatGPT or another hosting application may process prompts, selected files and outputs under its own terms, settings and retention policies. A cloud execution environment is not a promise of on-device processing. The kit's lack of telemetry does not change the host's behavior.
No personal data is required. Use synthetic or non-personal inputs, and do not put credentials, personal information or private research in public support reports. See the privacy notice for version 0.3.1.
Warranty and support
The software is supplied “as is,” without warranty, under the MIT License. Its warranty and liability provisions apply to the extent permitted by applicable law. Numerical results and record verification can be affected by software defects, numerical limits or unsuitable inputs.
Report reproducible software problems through GitHub Issues, using a small synthetic example. Issues are public and subject to GitHub's policies. There is no guaranteed response time or paid support service-level agreement. See the support guidance.
Versions and other platforms
Check the documentation and license shipped with the version you use. Changes to this page do not revoke rights already granted under the MIT License. The kit is maintained independently by Adem Vessell; availability in a host application's directory does not imply that host's endorsement or certification.