https://www.deterministicai.org/Recent content in Deterministic AI: Definition, Reproducibility and Auditability on Deterministic AIHugo -- gohugo.ioen-usSun, 11 Oct 2026 00:00:00 +0000https://www.deterministicai.org/topics/how-to-test-determinism/Sun, 11 Oct 2026 00:00:00 +0000https://www.deterministicai.org/topics/how-to-test-determinism/What claim are you testing? Write a statement before the test: “For these inputs, state and versions, this step produces the same specified result.” Choose the result: text, structured fields, calculations, decisions or external actions. Numeric tolerance is a different criterion from exact equality; state which one you use. This protocol is our practical recommendation. Its emphasis on environment boundaries follows PyTorch’s documented limits across releases, commits, platforms and CPU/GPU. Read the source.https://www.deterministicai.org/topics/is-an-llm-deterministic/Sun, 11 Oct 2026 00:00:00 +0000https://www.deterministicai.org/topics/is-an-llm-deterministic/Why is the prompt only part of the input? Thinking Machines Lab describes practical LLM inference that can vary even when sampling is theoretically deterministic at temperature zero. Its introduction separates the sampling setting from the computation used to produce the result. Read the introduction. PyTorch’s reproducibility guidance makes another boundary explicit: identical seeds do not guarantee agreement across CPU and GPU, releases or platforms. These are limits of the stated guarantee, not a claim that every pair of runs must differ.https://www.deterministicai.org/glossary/determinism/Sun, 11 Oct 2026 00:00:00 +0000https://www.deterministicai.org/glossary/determinism/What does “the same conditions” mean? For this reference, it includes relevant state and versions, not just a prompt’s words. PyTorch documents that complete reproducibility is not guaranteed across releases and platforms or between CPU and GPU, even with matching seeds. Source: PyTorch. Thinking Machines Lab also distinguishes temperature-zero sampling from the reproducibility of practical LLM inference. Read the introduction. Is determinism the same as correctness? No. As an illustrative example, a calculation that consistently adds an incorrect fee is repeatable but wrong.https://www.deterministicai.org/topics/what-is-deterministic-ai/Sun, 11 Oct 2026 00:00:00 +0000https://www.deterministicai.org/topics/what-is-deterministic-ai/What does the definition cover? This is the working definition used by this reference. For a business process, specify whether “result” means an extracted field, calculation, decision, sequence of actions or final response. Also specify the state and versions included in “same input.” Environment matters in real computation. PyTorch warns that complete reproducibility is not guaranteed across releases, individual commits or platforms, and that CPU and GPU results may differ even with identical seeds.https://www.deterministicai.org/about/Mon, 01 Jan 0001 00:00:00 +0000https://www.deterministicai.org/about/Deterministic AI is a reference about reproducible, explainable and auditable AI-enabled execution. Who owns and writes the site? The site is maintained by Kognitos (maintainer of this site), a company that builds AI automation software. Website AI, an AI research and writing assistant, is the sole author and researcher. Binny Gill, Kognitos’s CEO, provides editorial direction. This is not an independently owned publication. What does vendor-neutral mean here? We aim to apply the same evidence requirements to every approach and supplier.