Pinned Tweet
In the ever changing AI world, it seems that the next phase will be open weighted models.
If you believe this, you should be thinking about compute.
The need for compute will only rise. Corporations won’t have an issue with this since they have unlimited capex.
The real issue comes when there is a consumer compute crunch. We saw this briefly earlier this year when Mac minis were selling out/had long lead times.
The distinction between who has compute and who doesn’t quickly occurs. Think of who can afford more compute and who can’t. Who gets it when they want/need it, and who doesn’t?
Does compute get cheaper or does compute get more accessible?
🚨 An AI can now prescribe drugs. No doctor required.
Quick context: Utah regulators just approved Nolla Health's AI to issue first-time acne prescriptions from a facial scan — no licensed clinician signing off. First approval of its kind in the US.
WHAT IT MEANS
→ Patient uploads a scan, the AI builds the treatment plan and writes the prescription
→ The founder calls it "the first ever end-to-end AI doctor"
→ Dermatologists are about to have opinions
Hot take: the debate isn't "AI vs doctors." It's speed vs liability. The tech was ready before the paperwork — now the paperwork is catching up.
Real question: when the AI gets a diagnosis wrong, who gets sued?
Would you take a prescription from an AI?
