RAMDeck Pocket AI: Our Atlanta Home AI in Our Pocket in Madrid (39.1 tok/s, $0/mo)
RAMDeck is a palm-sized, plug-and-play hardware hub ($199 one-time, zero coding or terminal commands) that plugs into your home router in under 5 minutes and combines the RAM and GPUs of the everyday computers sitting around your house into one private AI brain. Our computers stayed plugged in on a desk in Atlanta while a regular Samsung Galaxy phone 4,370 miles across the Atlantic Ocean in Madrid, Spain chatted privately with a pooled 30B Qwen model over built-in Tailscale Remote at 39.1 tokens/sec—$0/month, zero cloud logging, and zero phone battery drain.
1. What RAMDeck Actually Is — And Why Phone AI Normally Forces a Lousy Choice
If you are hearing about RAMDeck for the first time, here is the simple version: big AI models like 30B and 70B Qwen, Llama, and DeepSeek are too big for one smartphone—and usually too big for any single laptop. So normally, phone AI forces two bad options:
Option 1: Pay $20 to $200/month forever to rent ChatGPT, Claude, or Gemini, sending every private business note, code snippet, and question through somebody else’s cloud servers. Option 2: Try cramming small local models directly onto your phone, watching your battery cook and bigger weights crash immediately with an Out-of-Memory (OOM) crash.
RAMDeck fixes both with hardware you already own: plug our tiny router hub in at home, connect the everyday Windows PCs, Macs, spare laptops, and Linux machines sitting in your house in under 5 minutes, and they merge into one private living-room cluster. Then our free iOS and Android app puts that Atlanta-style desk cluster in your pocket anywhere on Earth.
2. Live Simultaneous Proof: iPhone at Home (43.0 tok/s) & Samsung in Madrid 4,370 Miles Away (39.1 tok/s)
No synthetic mockups: below are side-by-side simultaneous phone captures hitting our pooled Qwen3-Coder-30B cluster sitting back on a desk in Atlanta—on iPhone at home over Home LAN (173 tokens at 43.0 tok/s with crisp native Markdown formatting) and on a Samsung Galaxy pulled out of a pocket 4,370 miles across the Atlantic in Madrid, Spain over encrypted Tailscale Remote (353 tokens at 39.1 tok/s). Crossing 4,370 miles kept over 90% of local home Wi-Fi speed.
Simultaneous Home LAN capture (173 tokens) with formatted Markdown titles and bullets at zero cloud cost.
Taken live in Madrid across the Atlantic: 353 tokens streamed at >90% local Wi-Fi speed with 0% phone battery burn.
3. Three Everyday Wins When Your Cluster Lives in Your Pocket
Automatic Home LAN ↔ Tailscale Failover (Zero Router Port Forwarding)
Open the RAMDeck app on home Wi-Fi and it speaks directly to your hub over Home LAN. Walk out the front door onto 5G cellular or airport Wi-Fi and it seamlessly switches to your private Tailscale Remote link—backed by our built-in 1-click Tailscale hub integration without exposing ports to the public internet.
0% Phone Battery & RAM Drain When Traveling
Your phone stays cool and sips battery because gigabytes of 30B–70B model weights remain parked across your pooled home desktop, workstation, or Mac. Plus, if you ever need emergency extra RAM back in the living room, the exact same iPhone or Android app can contribute mobile memory over Wi-Fi too.
Same Pooled Brain for Phone Chat and Desktop Coding Agents
Your mobile chat pocket app shares the same private hub runtime that feeds Cursor, Cline, Continue, and Open WebUI on your laptops—making it easy to combine plug-and-play home device pooling with zero-cloud privacy and zero monthly API invoices.
4. Cloud Phone Subscriptions vs. On-Device Phone Models vs. RAMDeck Pocket AI
| Mobile AI Setup | Monthly Fee | Privacy & Data Control | Model Power & Battery Impact |
|---|---|---|---|
| Cloud Subscription Apps | $20 – $200 / mo forever | Prompts routed through external third-party servers | Zero battery drain, strict paywalls & rate caps |
| On-Device Mini Phone Model | $0 | Private locally | Out-of-Memory (OOM) crash above tiny weights; drains phone battery & heats chassis |
| RAMDeck App + Pooled Home Hub | $0 / mo (One-time $199 Hub) | 100% private home cluster (Home LAN & Tailscale) | 39.1–43.0 tok/s on 30B Qwen with zero mobile thermal load away from home |
Frequently Asked Questions
How does RAMDeck let my phone run 30B Qwen, Llama, and DeepSeek models at 39.1 tok/s in Madrid 4,370 miles from our Atlanta desk?
RAMDeck is a turnkey plug-and-play consumer hardware hub that combines the RAM and GPUs of your everyday Windows PCs, Macs, Linux machines, and phones in under 5 minutes with zero terminal commands—so the heavy 30B and 70B weights stay loaded on your home cluster in Atlanta while the official iOS and Android apps stream replies back to Madrid over encrypted Tailscale Remote at 39.1 tok/s.
Does remote mobile chat over Tailscale drain my phone battery or RAM?
RAMDeck keeps your phone cool with zero battery or RAM drain: when chatting away from home over Tailscale Remote, your phone holds zero model gigabytes in local memory and performs zero heavy tensor compute—preventing Out-of-Memory (OOM) crashes and thermal throttling while your home hardware does the work.
Do I have to switch settings manually when leaving Home Wi-Fi?
RAMDeck switches automatically with zero manual settings: the iOS and Android companion apps use Home LAN when you are on home Wi-Fi and fail over to Tailscale Remote when you step onto cellular data or travel Wi-Fi.
How much does RAMDeck cost compared to $20-$200/month cloud chat subscriptions?
RAMDeck charges $0/month forever: both mobile companion apps are free on the Apple App Store and Google Play, powered by the one-time $199 RAMDeck Hardware Hub (134% Funded on Indiegogo, ships in January with 1 Free $29 Unplug-and-Walk Node USB).
Carry Your Home AI Cluster Everywhere ($0/mo)
Cancel monthly chat subscriptions and put private 30B–70B models in your pocket. Reserve the RAMDeck Hardware Hub for $199 (134% Funded on Indiegogo • ships in January • includes 1 Free $29 Official Unplug-and-Walk Node USB), or grab the companion apps free on mobile.
Trademark Notice: TradeMAV™ and RAMDeck™ are trademarks of TradeMAV LLC. All third-party product and model names—including Apple, iPhone, Android, Samsung Galaxy, Tailscale, Windows, Linux, Qwen, Llama, DeepSeek, Cursor, Cline, Continue.dev, and Open WebUI—are trademarks of their respective owners and are not affiliated with or endorsing TradeMAV LLC.
— The TradeMAV Team
