~/meetings / 2026-05-28
Meeting #13

> 2026-05-28

Whit demoed his full home server self-hosting setup — Tailscale, Jellyfin, Radarr, Sonar, and Orbstack — walking Lawrence through how it all fits together. The two also scoped out the Deli Cheese Inventory project and discussed collaboration workflow for the team. Wide-ranging discussion on AI privacy, autonomous agents, and surveillance tech.

2 attendees(Whit, Lawrence)
##topics covered
  1. 01.Home server demo — Whit screen-shared his setup running on a Mac Mini at home
  2. 02.Tailscale — free personal VPN mesh that lets you reach your home server from anywhere on any network
  3. 03.Media server stack: Sonar (TV), Radarr (movies), Prowlarr (torrent index), Jellyfin (streaming frontend with phone app)
  4. 04.Orbstack — lightweight Docker/container manager for Mac, running all services as containers
  5. 05.Railway as a hosting platform — $5/month, 6+ servers possible; restrictions include no piracy, no VMs, no bots
  6. 06.Servers explained — two aspects: compute (running applications) and storage (saving data)
  7. 07.DNS (Domain Name Servers) explained — how a domain like HEB.com resolves to a server's IP address
  8. 08.Static IP assignment — configuring your router to always give the same IP to your Mac so Tailscale stays connected
  9. 09.Hard drive storage as a failure point — Mac storage filling up kills running server processes
  10. 10.Mac M4 on-device AI potential — enough GPU to run large local models, cutting out cloud AI entirely
  11. 11.Deli Cheese Inventory project scoped — Christina (deli dept) does manual cheese inventory every few months; app would let her photograph labels, AI extracts name/weight/UPC, she reviews and saves
  12. 12.Two input workflows considered: one-by-one photo review, or batch photo upload with group review
  13. 13.Barcode scanning discussed as an alternative to full photo AI — UPC scan plugins exist
  14. 14.HEB API inaccessible (closed source) — can't use it to look up product names, so photo AI is necessary
  15. 15.Team contribution model discussed — figuring out each member's skills and how to divide work on a shared project
  16. 16.Collaboration-first mindset — keep scope small; the goal is to establish a shared codebase workflow, not just ship a tool
  17. 17.Claude Code voice mode — type /voice, hold spacebar to dictate directly into the terminal
  18. 18.OpenCode — open-source terminal AI coding assistant, an alternative to Claude Code that supports multiple models
  19. 19.Multi-model AI strategy — comparing answers across models, starting fresh conversations to get varied outputs
  20. 20.Autonomous AI agents — Cloudbot (formerly Klaba / Open Claw): agents with full permissions, no approval prompts
  21. 21.Multi-agent coordination pattern from Whit's Bart internship — shared message board so agents track each other's progress
  22. 22.AI privacy concerns — LLMs do sentiment analysis on everything you input; distrust of big tech owning the infrastructure
  23. 23.Anthropic noted as trying to differentiate on safety, but still backed by major surveillance-adjacent investors
  24. 24.Tutorial hell vs. just building — watching tutorials nonstop vs. diving in and figuring it out
  25. 25.Ben Jordan — content creator building 'AI Poison' to watermark music files so AI can't train on them correctly
  26. 26.Flock Safety — police license plate camera network; basically hackable, becoming a legal case
  27. 27.Synth ID — proposed government-enforced AI watermark standard to identify AI-generated content
  28. 28.AI-generated misinformation — how AI infographics spread as fact once shared out of context
##decisions & next steps
  • →Deli Cheese Inventory chosen as the next group collaboration project
  • →Scope kept deliberately small — use this project to establish a collaboration workflow for the team
  • →Ben Jordan's channel to be shared in Discord
##notes
None
← meeting #12meeting #14 →