Vibecodeando ando #013 Gasté $150,000 en Mac minis para esto
▶ Watch the video on YouTube
📜 Transcript (Text)
Automatic transcript · corrections or removal: DM @BandaWeb3
About this episode
Mexi unboxes five Mac minis and connects them in a ring with Thunderbolt cables to test local AI models. He explains why he prefers running his own models instead of depending on APIs, how an agent's memory works in OpenClaw, and what happened at the previous day's workshop, where each attendee had a server and a Minimax key. He closes with ideas for what to build first.
Chapters
- 00:00:00 Welcome and plan for the day
- 00:19:40 Unboxing the Mac minis
- 00:21:11 Agent memory and files
- 00:26:33 Thunderbolt ring connection
- 00:32:23 Recap of yesterday's workshop
- 00:37:49 Why run local models
- 00:49:46 What to build with these tools
- 00:55:22 Three ideas to start with
- 00:58:54 Shout-outs and wrap-up
Topics
Mac minis in a Thunderbolt ringLocal AI modelsPrivacy and API costsAgent memory in OpenClawWorkshop with virtual serversSaaS ideas to start withLosing the fear of the terminal