> ## Documentation Index > Fetch the complete documentation index at: https://invoca-5bd45748-mintlify-6c3474a6.mintlify.site/llms.txt > Use this file to discover all available pages before exploring further. # Reporting drift > What to do when Figma and code disagree — and how to change the system when you think code is wrong. ## Code is the source of truth Titan tokens are authored as DTCG JSON. Components and their intent live in code. Style Dictionary builds from that source and emits CSS, TypeScript, this documentation — and the Figma library. **Figma is an emission target, not an authority.** A value that exists only in Figma is not part of the system. This is a statement about **where values live**, not about **who decides**. Design authority over what the system *should* do is unchanged — see [Proposing a change](#proposing-a-change) below. What changed is that a decision is only real once it lands in source. ## Three kinds of disagreement "Drift" is used for three different situations, and only one of them has a mirror. Identify which you have before reaching for a fix, because the resolutions are not the same. | Between | What it means | Resolution | Recorded in | | ------------------------ | ------------------------------------------------ | ----------------------------- | ------------------------------------------------------------ | | **Code ↔ Figma** | The mirror is stale | Re-emit. Always. | This page | | **Specification ↔ code** | Two authored sources disagree | A decision, per token | [Divergences](/invoca-design-system/foundations/divergences) | | **Design intent ↔ code** | The docs say one thing, the package does another | Follow the register's verdict | [Divergences](/invoca-design-system/foundations/divergences) | **Only the first is a mirror problem.** The other two have no downstream side — both halves were authored by someone who meant it, so "re-emit" is not available and picking the better-looking one is not a resolution. Those get recorded and decided, not fixed in place. The largest example today: the naming specification and the shipped tokens overlap by only 54 of 244 names. See [TITAN-DIV-09](/invoca-design-system/foundations/divergences#titan-div-09). ## Code and Figma: drift has a direction When Figma and code disagree, the question is not "which one is right?" It is **"how far has the mirror drifted?"** The answer is always the same: ``` Code (DTCG JSON + component source) │ Style Dictionary ▼ CSS · TypeScript · Docs · Figma library ``` Resolution is **always** re-emission. Never edit the token to match the Figma file, and never hand-fix the Figma variable to match code — a hand-fix looks like a resolution and leaves the sync broken, so the same drift returns on the next publish with no record that anyone noticed. The instinct when two things disagree is to look at both and pick the better-looking one. That instinct is the thing this page exists to interrupt. "Figma looks more correct" is a reason to **propose a change to source**, not a reason to treat Figma as the answer. ## Reporting drift Name the token, component, or variant, and both values — the code value and what Figma shows. Screenshots help a human triage, but they are not the evidence. **A screenshot confirms a script ran; it does not confirm accuracy.** Two different problems with different fixes: * **Drift** — the value exists in code and Figma shows something else. The mirror is stale. Re-emit. * **Gap** — the value exists in *neither*, and what renders has no token behind it at all. Not drift. It needs a token added to source. See [Token workflow](/invoca-design-system/contribute/token-workflow). Gaps are the more common finding, and they are invisible without a systematic diff. An audit run this way on Titan surfaced roughly 20 discrepancies, none visible by eye. Include the token or component name, the code value, the Figma value, and which tier and axis it touches — primitive/semantic/component, and whether a mode is involved (light/dark, comfortable/compact). The fix runs the sync. If the sync produces the wrong result, the sync is the bug. ## Proposing a change **Reporting drift and proposing a change are different actions.** Confusing them is how "code is the source of truth" gets misheard as "designers do not get a say." | You think… | That is | Do this | | ------------------------------------------ | --------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------- | | Figma shows something different from code | **Drift** | Report it. Resolution is re-emission. | | The value in code is the wrong value | **A change proposal** | [Propose it](/invoca-design-system/contribute/propose-a-component). It gets decided, then it lands in source. | | The value I need does not exist anywhere | **A token gap** | [Token workflow](/invoca-design-system/contribute/token-workflow). Do not work around it with a literal. | | The documentation and the package disagree | **A divergence** | Check [Divergences](/invoca-design-system/foundations/divergences) for a `TITAN-DIV-`. If it is not recorded, report it — do not pick a side. | | Nothing anywhere decides this | **An open decision** | Check [Open decisions](/invoca-design-system/foundations/open-decisions). Cite the `TITAN-GAP-`; never resolve it by inference. | A designer changing a value in Figma is not authoring. A designer **deciding** a value absolutely is — the decision simply has to land in source to become real, the same way an engineer's decision does. Nobody's judgment is downstream here. Only the file format is. ## What not to do **Do not hand-edit generated output.** Not the emitted CSS, not the TypeScript, not the files under `snippets/generated/`. A hand-edit does not fix anything; it creates a value that disagrees with source and survives until someone notices. **Do not resolve drift with a local override.** An override is invisible to the audit, to Figma, and to every other consumer. It is how a design system stops being one. **Do not resolve a divergence by editing the documentation to match the code.** That is the non-mirror case, and quietly rewriting the intent destroys the decision. Record it in [Divergences](/invoca-design-system/foundations/divergences) instead. **Do not treat the legacy library as evidence.** The "IDS – Core Components" Figma file may be referenced for *design intent* — it carries real direction about how things were meant to feel — but it is not a source for values, for variant models, or for Code Connect mappings. It is history, not input. ## Related * [Token workflow](/invoca-design-system/contribute/token-workflow) — how a missing value becomes a token * [Versioning & deprecation](/invoca-design-system/contribute/versioning-and-deprecation) — what happens when one is removed * [Divergences](/invoca-design-system/foundations/divergences) — where code and intent disagree, and which to follow * [Open decisions](/invoca-design-system/foundations/open-decisions) — what has not been decided ## The health metric Titan tracks **generated-to-merged delta**: how much a human changes what the tools generate before it merges. A rising delta does not mean generation got worse. It means the mirror or the Code Connect mappings went stale — which is drift, measured before anyone reports it. That makes it the one number worth watching, and it is why documentation coverage is *not* the metric. A fully documented system whose mirror has drifted produces confidently wrong output.