Matt WolfeFutureTools.io · Founder of Future Tools | Host of Next Wave Podcast

Unpopular opinion: you don't need to ask ChatGPT everything

11 Jun 2026 · 1 min

Watch on YouTube

1:13

In this monologue, Matt Wolfe discusses Nvidia's recent announcements, focusing on the RTX Spark chip—a hybrid GPU/CPU with up to 128 GB of unified compute. He explains how this hardware makes it feasible to run sizable, sophisticated language models directly on personal devices, moving inference workloads away from cloud services. Wolfe highlights the advantages of local inference: reduced reliance on cloud resources for routine tasks, enhanced privacy because prompts never leave the device, and the ability to operate offline (e.g., on a plane). He argues that many everyday LLM applications can be handled by smaller, on‑device models, reserving cloud‑based, compute‑intensive models for truly demanding tasks.

Chapters

Nvidia made a series of announcements. The biggest, in my opinion, is the RTX Spark. This is a new chip that's both essentially a GPU and a CPU in one, and the current version has up to 128GB of unified compute, meaning you could run some pretty large, pretty smart LLMs locally on your device. And the whole idea here, they're trying to shift more and more of the inference, the stuff that happens after you give AI a prompt, directly onto your computer. So these bigger, more compute-intensive models that you're sending stuff to the cloud for, we can start saving those for the things that actually need that level of intelligence and let most general things run locally on your computer. That's what these companies are starting to push more into. If you're running local models, you don't have to send them to the cloud where they could be possibly training on your models or getting access to private details. You can also run them completely offline. So if you're like on a plane without internet access and you want to use a solid LLM, you don't even need to be connected to the internet to run AI models anymore. And a lot of what people use large language models for these days is fairly simple things that, for lack of a better word, the dumber models, the smaller models can actually handle for you.