Deterministic calculation is a calculation that gives the same answer from the same inputs, every time. Wikipedia defines a deterministic algorithm as one that, "given a particular input, will always produce the same output". Databases, accounting systems and spreadsheets work this way. Run last month's report again on the same data and the totals match.
Why it matters now
Large language models don't reliably work this way. In a 2025 test by Thinking Machines Lab, an AI research company, one prompt run 1,000 times at temperature zero produced "80 unique completions". That is fine for drafting a letter. It isn't fine for a stumpage total, a margin or an invoice. So a reliable AI system uses a model to understand the question and a deterministic tool to compute the answer.
Deterministic is not the same as exact
Deterministic code can still disagree with another system by a cent. Spreadsheets and most programming languages store numbers in binary floating point, and the Python documentation explains that "most decimal fractions cannot be represented exactly as binary fractions". The trouble shows up at rounding. A value that looks like exactly half a cent may be stored slightly below it, so it rounds down. The documentation's own example: "round(2.675, 2) gives 2.67 instead of the expected 2.68".
The fix is exact decimal storage, or whole cents, and one rounding rule: round each line or round the total, to how many places, and which way. The Python documentation recommends decimal arithmetic "suitable for accounting applications". In our experience, most cent differences between systems come from rules that differ, not from storage.
In a timber business
Every reported figure should be reproducible: the same tickets, the same rates, the same total, whoever runs it and whenever. When an AI assistant reports a volume or a cost, the number should come from a calculation you can rerun.
What it isn't
Reproducible isn't the same as right. A calculation that uses last year's rate will repeat that mistake on every run, but it can be traced and fixed.
How Quarri works explains the platform as a layer over existing systems, not a migration.
Sources
- Wikipedia, "Deterministic algorithm": en.wikipedia.org
- Python Software Foundation, "Floating-Point Arithmetic: Issues and Limitations", Python documentation, accessed 27 September 2026: docs.python.org
- Python Software Foundation, "Built-in Functions: round()", Python documentation, accessed 27 September 2026: docs.python.org
- Thinking Machines Lab, "Defeating Nondeterminism in LLM Inference", 2025: thinkingmachines.ai
Quarri is an AI-native data platform for the timber supply chain. It connects buying, production, sales and inventory for forest management, sawmill, wood products and pulp, paper and packaging operators.