Topic: Local Ai

3 chapters in Aug – Oct 2026 · See all time

Off The Rails
Episode 1915 • • 1:19:18 - 1:23:36

1915: Off The Rails

Grokbot, Banking APIs, and Local AI Models

The group discusses the capabilities of AI bots like Grokbot to manage personal finances and shopping through APIs and services like Plaid. Jason Calacanis predicts a future where "sovereign" AI agents run locally on computers with high RAM, bypassing the privacy and liability issues of cloud-based models. They explore how these bots can find better deals on products and even handle bank transfers in read-only or active modes.

Trust The Plan
Episode 1900 • • 1:20:29 - 1:24:09

1900: Trust The Plan

Local AI Hardware, Asus DGX 128GB Machine

Adam Curry describes acquiring an Asus-branded DGX machine with 128GB of RAM to run AI models locally, avoiding monthly subscription fees and cloud outages. The machine, which has doubled in price due to RAM costs, allows for high-performance tasks without relying on commercial APIs. The discussion touches on the measurement of "compute" in flops versus power consumption.

asusdgxblackwell chiplocal airam pricescomputeteraflops
XY You're Out
Episode 1895 • • 1:57:55 - 2:05:21

1895: XY You're Out

Data Center Construction, Security and AI Infrastructure

A "boots on the ground" report from Sir Darius Unity describes the high-security environment of modern data center construction. Workers are often required to surrender phones to prevent unauthorized photos of the sensitive infrastructure. The report details the 24/7 operations, armed security, and the massive financial losses incurred during even a minute of downtime. The hosts discuss whether the current investment in data centers for AI is a bubble.

data centerssir darius unityunion electricianssecurityaiinfrastructurelocal 26

For AI agents: this site has a plain-text version made for you. This page in markdown: https://search.noagendashow.net/topic/local-ai.md. Search every episode transcript: https://search.noagendashow.net/search.md?q={query}. How to use the site, with timestamped links to quote: https://search.noagendashow.net/llms.txt.