Your Mac can run its own private AI — no subscription, no internet. One problem: it reads big stuff slowly, and it gets forgetful in long chats.
Meanwhile your iPhone sits there, bored. 
Good news: it can help — with the cable already in your drawer. 
The deal, in plain words:
Your Mac’s AI reads big files up to 44% faster (the back half of the work runs on the phone’s graphics chip)
It remembers ~3× more of long chats — the oldest part quietly moves to your phone
Answers come out exactly the same — checked word-for-word, not faked
Cost: $0 — free tool, free models, one cable
And it gets crazier 
- An old 8GB MacBook runs the big brain too — the phone carries half of it (recipe)
- Two phones = up to 262k memory — the chat never has to end (how)
- In testing: a ~180B model spread across a whole house — Mac + 2 PCs + 2 phones (roadmap)
- Works on iPhone 15 Pro → 18 Pro and every M1 iPad
And the other devices? They can help 
A Windows PC or Ubuntu box joins as a helper today (Ubuntu guide — brand new, first results wanted 
). Android and even an Apple TV are on the wish list.
Set it up in 3 steps 
1. Mac: paste this one line (free, installs everything):
curl -fsSL https://raw.githubusercontent.com/StayLameBro/backburner/main/install.sh | bash
2. iPhone: free app via AltStore — no developer account. (AltStore 2.2+, not SideStore — SideStore breaks the magic.)
3. Plug in. Done. 
🗄️ Fine print & secret perks (1 minute, worth it)
Any fast USB-C cable works — the one in the iPhone box is USB 2, too slow
First run: allow Local Network when iOS asks — until then it fails silently
After every Mac reboot: sudo sysctl iogpu.wired_limit_mb=20480
Keep the app open on screen while it works — an iPhone rule, not the tool’s fault
16–18GB Macs: run a lighter model, recipe in the README
Keep the app updated (0.0.3+ fixed an early Wi-Fi hole); the phone answers the cable only
Secret perks: known chats restart in seconds, a third phone can join over Wi-Fi (doc), per-device speed table in the README
For the spec-hungry: the iPhone 17 Pro’s new matrix units make its half 2.4× faster; layer splits per device live in the README table (17 Pro: 40, older Pros: 52, M1 iPad: 52)
Free models: bartowski’s GGUF shelf
HQ: github.com/StayLameBro/backburner
Why should insiders have all the fun? The cable is already in your drawer. 
!