Agentic AI, Self-Improving AI Systems, Marcel Builds An Agent Network and Raccoon & Sovereign AI

Agentic AI Goes Enterprise: NVIDIA’s NeMo-CLAW, Small Local Models, and a Secure Alternative with Raccoon GPT. (https://raccoongpt.ca)

Hashtag Trending would like to thank Meter for their support in bringing you this podcast. Meter delivers a complete networking stack, wired, wireless and cellular in one integrated solution that's built for performance and scale. You can find them at Meter.com/htt

On Project Synapse, Jim Love, Marcel G, John Pinard, and guest Tony Kaye from Raccoon GPT  discuss a week of major AI developments, including: Jensen Huang’s keynote highlighting NVIDIA’s NeMo-CLAW security layer and an enterprise-focused, more secure approach to agent tools like Open CLAW. 

They explore why agentic AI matters—moving from chat to automated actions—along with the risks of insecurity and shadow AI. The panel also covers the emergence of smaller models intended to run locally for secure workflows, Andrej Karpathy’s auto-research self-learning experiment showing an 11% improvement, and Jim’s week-long build of “Nexus,” a personal secure agent with memory, journaling, Telegram and web access, voice via Whisper, digests, and a dashboard. Tony explains Raccoon GPT’s Canada-based, privacy-first architecture for regulated organizations.

00:00 Show Intro Sponsor
00:18 Meet Tony K
00:57 Lightning Round Setup
01:51 Nvidia Nemo Claw
04:10 Why Agents Matter
05:36 Security And Shadow AI
12:04 Small Local Models
13:35 Open Source Model Shift
19:19 Self Learning Breakthroughs
26:07 Marcel Builds Nexus
34:01 Cutting Telegram Costs
34:58 Nexus Personal Workflow
37:24 Open Source Chaos
40:12 Raccoon Secure AI
46:34 Data Privacy Reality
50:30 Enterprise Agent Foundations
51:38 Models Architecture Costs
56:27 Sovereignty and Control
01:02:02 Agents Remake the Web
01:06:01 Wrap Up and Sponsor