Perplexity AI Meta Manus Desktop App, Anthropic Enterprise Lead, OpenAI AWS Deal
Mar 18, 2026
Discussion of a new desktop AI agent that integrates with local files and apps and what that means for agents becoming an OS layer. Exploration of a tool cutting real‑time data center energy spikes for AI workloads. Look at visual memory systems for wearables and robots to recall real-world experiences. Review of shifting enterprise AI spending and a major AI provider's government cloud deal.
AI Snips
Chapters
Transcript
Episode notes
Agents Moving From Cloud To Local OS Layer
- Meta's Manus desktop brings AI agents from cloud to local machines to act on files and run apps.
- Jaeden Schafer highlights this as a shift toward agents becoming an OS layer that can build software locally.
Weigh Privacy When Running Local AI Agents
- Expect greater privacy and security trade-offs as local agents access files and run apps on users' machines.
- Jaeden Schafer warns Meta's privacy track record means some users will hesitate to give Manus deep local access.
Energy Is The Hidden Bottleneck For AI
- Niv AI focuses on real-time power optimization to smooth GPU workload spikes in AI data centers.
- Jaeden Schafer frames energy as a core bottleneck, saying firms that squeeze more output from same hardware gain a big advantage.