Operating Principles¶
These are the practical rules for how the company runs day to day, across all ventures. They are more concrete than the mission/vision and are meant to be directly actionable.
Evidence-based decisions¶
Decisions — product, pricing, prioritization, architecture — should be grounded in evidence where evidence is available, and clearly labeled as assumption or hypothesis where it is not. See docs/company/research-and-evidence-principles.md for the full standard on how evidence is categorized and how fabrication is prohibited. In practice: don't state a guess as a fact, and don't let a confident-sounding guess quietly become "known" over time just because it was repeated.
Documentation as a first-class artifact¶
Documentation is not an afterthought bolted onto finished work — it is part of the work. A decision, standard, or product specification that exists only in someone's head or in a chat transcript is not considered done. This applies equally to human and AI-authored work: see AGENTS.md for how this is enforced for agents.
Git as source of truth¶
This Git repository is canonical for approved company documentation. Notion and other tools are used for day-to-day management and roadmap tracking, but they do not override what is recorded here. If a Notion page and this repository disagree, this repository controls until it is updated. See the root README.md for more on the Git/Notion split.
Proportionality¶
Process should match the size and risk of the thing it governs. A small internal script does not need the same rigor as a customer-facing pricing claim. A single-tab spreadsheet template does not need the same specification depth as a multi-user application. Applying enterprise-grade process to small tools wastes time without reducing real risk — but skipping process on genuinely risky work (data handling, financial claims, health/food-safety claims) is not acceptable regardless of company size. Use judgment, and when unsure, err toward more documentation for anything that would be expensive to get wrong.
Transparency of assumptions¶
Every plan, specification, or decision that rests on an assumption should say so, explicitly. Hidden assumptions are how small early-stage mistakes become expensive later mistakes. Unresolved assumptions and open questions are tracked centrally:
This applies to both humans and AI agents working in this repository. An AI agent that has to guess at something the task didn't specify must say what it assumed, not just proceed silently.
How these principles interact¶
These principles are meant to reinforce each other: evidence-based decisions require documentation to be recorded and trusted; documentation only stays trustworthy if it lives in one canonical place (Git); proportionate process keeps documentation sustainable rather than becoming its own burden; and transparent assumptions keep the whole system honest about what is actually known versus guessed.