The $2,000 Computer That Skips the Cloud

Most conversations about AI in medicine assume a data center somewhere else is doing the thinking. This week’s Signal & Symptoms starts by throwing that assumption out. Dr. Junaid Kalia is running a genuinely capable AI model on a $2,000 Mac on his own desk — no cloud, no per-token bill, no patient data leaving the building — and he’s asking the rest of us why we’re still renting.

No guest this episode. Just Junaid with co-hosts Ed Marx and Dr. Harvey Castro, talking through what “local AI” actually means for clinicians — and why it might matter more than the next frontier model announcement.

• Local vs. cloud, in plain language. Edge (your phone), near-edge (your laptop), far-edge (a private cloud) — and why where a model runs is now a clinical and privacy decision, not just an IT one.

• What you can actually run today. The “onslaught” of free, open-weight models — small router/tool models, mid-size workhorses, and frontier-class ones — on hardware you can buy off the shelf.

• The catch nobody wants to say out loud. The best cheap models right now are largely Chinese, which turns a technical choice into a question about standards, security, and national competitiveness.

• The medicine of it. Why a model is a black box the FDA’s Software Bill of Materials can’t fully itemize — and why a rural clinic with no road may be local AI’s best use case.

The bet in this episode isn’t that local AI replaces the cloud tomorrow. It’s that clinicians who understand the choice — own the model or rent it, keep the data or ship it — will make better decisions for their patients than those who assume there’s only one option. As Harvey put it in his closing: learn to use one of these tools today, even for something small.

“I’m getting a $2,000 Mac giving me 80 tokens a second. I can get ten of them — much better than giving money to Azure.”

Dr. Junaid Kalia, Neurocritical Care Specialist, Founder of Savelife.AI™

Dr. Harvey Castro, ER Physician, #DrGPT™

Edward Marx, CEO, Advisor