I can’t take credit for cracking the code, but I will take credit for this little app that makes it easy to use. Apfel is a wonderful OSS code library that unlocks the Neural Engine on Apple’s M-series chips, which normally is locked-down for their own use in Writing Tools, Apple Intelligence, Xcode’s assistant and more. However, it’s a CLI utility so requires hopping into a terminal to use, so I wrote Grapfel, to wrap around Apfel and give you a friendly way to hit that sweet Neural Engine.
Grapfel clipboard formatting options – Markdown, code or plain text
You can copy output as plain text, Markdown or code blocks (the Neural Engine is surprisingly good at Apple-based technology and coding standards, FYI), and tweak the algorithm’s temperature and system prompt to your liking. Grapfel also supports attaching small text files to the conversation for limited context window tweaking (but the API doesn’t allow very large files – I’m working on a workaround for this so you can upload bigger papers, research, etc).
No fees, no subscriptions, no security issues (it’s running locally on your own hardware) – Apfel opens up the fun and Grapfel lets you get your hands on it easily. It only works on Silicon-based Macs (M-series chips) as that’s where the Neural Engine lies, and offered as-is for now. If you like it and want more features/changes or run into bugs, let me know by logging an issue on GitHub with:
- the specific model of Mac, and version of MacOS you’re running
- what you expected to happen, and what actually happened
- any prompts you were using or working with when the issue occurred
- any crash logs that were generated,or error messages displayed
Have fun playing with Apple’s hardware-based AI model! I will say that it absolutely sucks at modern/pop culture references, but is great with general advice and AMAZING at Apple technology & development practices. But you get what you pay for 🙂