Episode publish date
January 20, 2026 7:10 PM (UTC)
Last edit date
Feb 22, 2026 4:26 PM
Last snip date
February 22, 2026 4:19 PM (UTC)
Last sync date
February 22, 2026 4:25 PM (UTC)
Show
Practical AI
Show notes link
Snips
2
Warning
⚠️ Any content within the episode information, snip blocks might be updated or overwritten by Snipd in a future sync. Add your edits or additional notes outside these blocks to keep them safe.
‣
Your snips
‣
[02:15] From Meta And Roblox To RINX
‣
[06:43] Black-Box Guardrails Are Insufficient
‣
[09:26] Begin With Risk Categories
‣
[11:48] Interpretability Unlocks Runtime Control
‣
[19:54] Catch Bad Outputs Before They Finish
‣
[24:55] Add Safety As A Separate Layer
‣
[27:58] Internal Signals Are Much Cheaper
‣
[36:55] AI Security: Defense in Depth
‣