Episode Link
https://share.snipd.com/episode/b3f17c10-c8f6-484d-aa9b-4efd2e831def
Episode publish date
December 23, 2025 1:06 AM (UTC)
Last edit date
Jan 17, 2026 5:53 PM
Last snip date
January 17, 2026 5:52 PM (UTC)
Last sync date
January 17, 2026 5:52 PM (UTC)
Show
This Day in AI Podcast
Show notes link
https://share.transistor.fm/s/325fe844
Snips
20
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
‣
[01:06] Shipping Models From The Beach
‣
[02:08] Flash-Level Models Punch Above Weight
‣
[10:45] Image Models Compete On Reliability
‣
[14:20] Safety Filters Seem Looser Now
‣
[21:26] Use Firecrawl For Reliable Research
‣
[28:22] Deep Research Agents Structure Discovery
‣
[30:33] Combined Research Method
‣
[35:03] Skills Reduce Token Bloat
‣
[40:24] Codify Procedures As Skills
‣
[47:43] Load Heavy Context Only When Needed
‣
[51:52] Try Skills Early And Share Them
‣
[53:20] Mid-Range Reliability Beats Frontier Volatility
‣
[02:08] Episode summary — Gemini 3 Flash, GPT-Image-1.5, Skills vs MCPs, and 2025 recap
‣
[01:03:04] Hosts' Best Model Picks
‣
[01:04:00] Most Disappointing Major Releases
‣
[01:05:13] Skills Will Spread, Execution May Decouple
‣
[35:03] What "skills" are (concise explainer)
‣
[01:07:25] Run Skills Outside Model Providers
‣
[01:09:31] Agentic Workflows Will Change Jobs
‣