---
title: "Platform Examples"
description: "Minimal target and shell examples for macOS CPU, Linux CPU/CUDA, and Windows CPU/CUDA."
source: https://scrollcase.dev/v2/guides/platform-examples
deprecated: true
schema-version: 2
current-schema-version: 3
current: https://scrollcase.dev/guides/platform-examples
---

> **DEPRECATED.** This documents Scrollcase box format schema version 2, superseded by schema version 3 and no longer maintained. A version 3 verifier refuses a version 2 box by name rather than reading it, so nothing here describes the current release. The page that supersedes this one is https://scrollcase.dev/guides/platform-examples

# Platform Examples

Every box is built natively for one exact target. These snippets show target declarations and shell
syntax; they do not claim that a particular scroll or scientific workload has been validated.
Replace the example `<boxId>/<targetId>` references and use the exact `pixiVersion` pinned by each
scroll.

  
**macOS CPU**

```jsonc
"target": { "platform": "macos", "arch": "aarch64", "accelerator": "cpu" }
```

```sh
scrollcase doctor --scroll my-box/macos-aarch64-cpu
scrollcase lock my-box/macos-aarch64-cpu
scrollcase audit my-box/macos-aarch64-cpu
scrollcase build my-box/macos-aarch64-cpu
```

The interpreter is `venv/bin/python`; the target ID is `macos-aarch64-cpu`.

  
  
**Linux CPU**

```jsonc
"target": { "platform": "linux", "arch": "x86_64", "accelerator": "cpu" }
```

```sh
scrollcase doctor --scroll my-box/linux-x86_64-cpu
scrollcase lock my-box/linux-x86_64-cpu
scrollcase audit my-box/linux-x86_64-cpu
scrollcase build my-box/linux-x86_64-cpu
```

The interpreter is `venv/bin/python`; the target ID is `linux-x86_64-cpu`.

  
  
**Linux CUDA**

```jsonc
"target": {
  "platform": "linux",
  "arch": "x86_64",
  "accelerator": "cuda",
  "cudaVersion": "12.4"
}
```

```sh
scrollcase doctor --scroll my-box/linux-x86_64-cuda12.4
scrollcase lock my-box/linux-x86_64-cuda12.4
scrollcase audit my-box/linux-x86_64-cuda12.4
scrollcase build my-box/linux-x86_64-cuda12.4
```

`12.4` is an example ABI. The generic target suffix is `cuda<major.minor>`. A successful native
build proves packaging and declared gates, not scientific parity with another accelerator.

  
  
**Windows CPU**

```jsonc
"target": { "platform": "windows", "arch": "x86_64", "accelerator": "cpu" }
```

```powershell
scrollcase doctor --scroll my-box/windows-x86_64-cpu
scrollcase lock my-box/windows-x86_64-cpu
scrollcase audit my-box/windows-x86_64-cpu
scrollcase build my-box/windows-x86_64-cpu
```

The interpreter is `venv/python.exe`; the target ID is `windows-x86_64-cpu`.

  
  
**Windows CUDA**

```jsonc
"target": {
  "platform": "windows",
  "arch": "x86_64",
  "accelerator": "cuda",
  "cudaVersion": "12.4"
}
```

```powershell
scrollcase doctor --scroll my-box/windows-x86_64-cuda12.4
scrollcase lock my-box/windows-x86_64-cuda12.4
scrollcase audit my-box/windows-x86_64-cuda12.4
scrollcase build my-box/windows-x86_64-cuda12.4
```

Windows CUDA is defined by the target contract and is buildable on a matching native host. This
example is not a claim that a specific box, GPU, driver, or scientific workload is supported.

  

`verify --self-test` also requires the matching native OS and architecture. CPU and CUDA boxes
share neither identity nor implied compatibility; build and verify each target independently.