diff --git a/sandbox/README.md b/sandbox/README.md
new file mode 100644
index 00000000..517dcff3
--- /dev/null
+++ b/sandbox/README.md
@@ -0,0 +1,135 @@
+# Sandbox — Pre-Release & Experimental Code
+
+This directory is the home for experimental features, prototypes, algorithm
+explorations, and hardware integration tests that are not yet ready for
+production. If you're trying something new, this is where it lives until it's
+validated and promoted.
+
+## Why Sandbox Exists
+
+Instead of creating a new repo for every experiment — which leads to orphaned
+repos, inconsistent naming, and invisible work — we keep pre-release code in the
+platform repos, where it belongs. This makes experiments discoverable,
+reviewable, and easy to promote when they're ready.
+
+## Rules
+
+### 1. Isolation
+
+Sandbox code is **never imported by production code**. The main build system
+explicitly excludes this directory. No production module may reference anything
+in `sandbox/`.
+
+If your experiment depends on production code, import the installed package
+(e.g., `import openlifu`, `#include` from installed headers). Never use relative
+paths back into `src/` or `Core/`.
+
+### 2. Self-Contained
+
+Each experiment is a standalone folder. It must include everything someone needs
+to understand and run it: its own README, dependencies, sample data (or
+instructions to get it), and run instructions. Assume the next person has never
+seen your code.
+
+### 3. Ownership
+
+Every experiment folder **must** have a `README.md` with the following header:
+
+```markdown
+# [Experiment Name]
+
+| Field                 | Value                                               |
+| --------------------- | --------------------------------------------------- |
+| **Status**            | `prototype` / `testing` / `validating` / `archived` |
+| **Owner**             | [Your name]                                         |
+| **Created**           | [Date]                                              |
+| **Target graduation** | [Quarter/date or "exploratory"]                     |
+
+## Description
+
+[One paragraph: what this is, what problem it solves, why it matters.]
+```
+
+### 4. Status Definitions
+
+| Status       | Meaning                                                               | What happens next                   |
+| ------------ | --------------------------------------------------------------------- | ----------------------------------- |
+| `prototype`  | Early exploration, proof-of-concept. May not run cleanly.             | Continue development or archive.    |
+| `testing`    | Core functionality works. Being tested against real data or hardware. | Move to validation or archive.      |
+| `validating` | Feature-complete and under review for production promotion.           | Open a graduation PR or archive.    |
+| `archived`   | Experiment concluded — either graduated or abandoned.                 | Move to `_archived/` with a reason. |
+
+### 5. Graduation Path
+
+When an experiment is ready for production:
+
+1. Open a PR that moves the code from `sandbox/[experiment]/` into the
+   appropriate production directory (`src/`, `Core/Src/`, `processing/`,
+   `Modules/`, etc.)
+2. Add corresponding tests
+3. Remove the sandbox folder in the same PR
+4. Reference the original sandbox README in the PR description for context
+5. Get a code review from someone who didn't write the experiment
+
+The PR is the gate — sandbox code doesn't sneak into production. It gets a
+proper review, tests, and documentation before it graduates.
+
+### 6. Archival
+
+Experiments that don't graduate get moved to `sandbox/_archived/` with an
+updated README:
+
+```markdown
+| **Status** | `archived` | | **Archived** | [Date] | | **Reason** | [Why —
+e.g., "approach superseded by X", "hardware target canceled", "merged into
+production as Y"] |
+```
+
+Archived experiments are kept for historical reference. They may contain useful
+ideas, data, or lessons even if the code itself didn't ship.
+
+### 7. CI Treatment
+
+CI runs a **separate, lightweight job** on `sandbox/` — linting, type checking,
+and basic syntax validation where applicable. Sandbox CI failures are **visible
+but non-blocking**: they do not gate merges to `main` or production releases.
+
+This keeps quality visible without penalizing experimentation.
+
+### 8. No New Repos for Experiments
+
+If it's a prototype, an algorithm experiment, a new sensor integration, or a
+hardware test script — it goes here. New repositories are created only for
+genuinely new standalone components: a new firmware target for a new MCU, a new
+PCB design, or a new standalone tool with its own release cycle.
+
+When in doubt, start in a sandbox. You can always extract to a new repo later if
+it grows into something that justifies its own lifecycle.
+
+## Directory Layout
+
+```
+sandbox/
+├── README.md                    # This file
+├── [experiment-name]/           # One folder per experiment
+│   ├── README.md                # Status, owner, description (required)
+│   ├── [code, data, configs]    # Whatever the experiment needs
+│   └── ...
+└── _archived/                   # Concluded experiments (kept for reference)
+    └── [old-experiment]/
+        └── README.md            # Must include archive reason
+```
+
+### Naming Conventions
+
+- Folder names: **lowercase-with-hyphens** (e.g., `multi-element-beamforming`,
+  `speckle-contrast-v2`)
+- Be descriptive: `new-algo` is bad, `real-time-dcs-classification` is good
+- Prefix with the component area if it helps: `ui-3d-overlay`,
+  `fw-power-sequencing`
+
+## Questions?
+
+If you're unsure whether something belongs in a sandbox, a new repo, or a
+feature branch, ask in the team channel or open a discussion. The goal is to
+keep experiments visible and the org clean — not to create bureaucracy.