TwoMillionKit: Private Cloud Compute in FoundationModels without an entitlement (github.com)

Apple has introduced the ability for third-party apps to utilize its Private Cloud Compute language models in macOS 27. However, this feature is restricted to developers enrolled in the Small Business Program whose apps haven't exceeded 2 million downloads. This limitation poses a challenge for developers with popular apps, as they cannot access these advanced AI capabilities, putting them at a competitive disadvantage. Moreover, issues surrounding Apple’s reporting on download counts complicate the situation, as some developers are unable to verify their eligibility for this feature. In response to these limitations, a developer has created a workaround by utilizing the `fm` command-line tool included in macOS 27. This implementation allows Mac apps, particularly those distributed outside the Mac App Store, to leverage Private Cloud Compute without the need for special entitlements. While this method opens up new possibilities for small developers to integrate AI features into their applications, it does come with the caveat that sandboxed apps, which are required for App Store distribution, are not compatible. This development underscores both the creative ingenuity within the developer community and the ongoing challenges posed by corporate restrictions in accessing powerful AI tools.