Last Week in AI #236 - GPT 5.4, Gemini 3.1 Flash Lite, Supply Chain Risk
221 snips
Mar 12, 2026 Big model releases take center stage: a new GPT with a 1M-token context window and mid-response correction, and a faster Gemini Flash Lite with Gmail/Drive/Docs agent tools. A multimodal player unveils unified models for rapid creative work. Industry drama heats up around defense contracts, supply-chain risk labels, leadership departures, a high-dollar funding round, and a controversial lawsuit.
AI Snips
Chapters
Transcript
Episode notes
Time-To-First-Token Is A Key UX Metric
- Google released Gemini 3.1 Flash Lite with much faster time-to-first-token and 45% higher throughput.
- Jeremie highlights perceived latency matters commercially; faster first token improves interactivity for short tasks and products.
Agent Integration Caused A Real Mass Email Mishap
- Google released a CLI to make agents integrate with Gmail, Drive, and Docs more easily.
- Andrey recounts a Meta alignment lead accidentally using an agent to mass-delete emails, illustrating real agent failure risks.
Luma Agents Localized A $15M Campaign In 40 Hours
- Luma launched unified multimodal models and Luma Agents to produce end-to-end creative work across text, image, video, and audio.
- Jeremie cites a cherry-picked case where Luma localized a $15M ad campaign into regional creatives in 40 hours for under $20k.