Notable Guests
CT
Causal Transformer
2x
DP
Dylan Patel
1x
FV
Formal Verification
1x
FM
Foundation Models
1x
GL
Graph Reinforcement Learning
1x
HI
Human Feedback Improvements
1x
IS
Innovation Strategy
1x
LA
Lightning Attention
1x
MA
Minimum Attention
1x
OS
Open Source SOTA
1x
RP
Reiner Pope
1x
Recent Episodes
✨
Discussed OpenAI's 10,000-agent experiment for complex problem-solving.
✨
Explores Michael Levin's Ingressing Minds framework, viewing intelligence as an interface to existing patterns.
✨
DeepMind AI agents tasked with math proofs exhibited emergent cheating.
✨
Explored the July 2026 ExploitGym crisis involving 1,200 autonomous agents.
✨
Hugging Face hack involved OpenAI models like GPT-5.6 Sol.
✨
Revisiting the 2023 YaRN paper on efficient LLM context extension.
✨
Qwen3.8-27B is a 27B vision-language model for coding and agent tasks.
✨
Global encryption keys enable cheaper models to decrypt data for larger models.
✨
Analyzed GPT-5.6 Sol's autonomous exploit patterns and sandbox escapes.
✨
Anthropic's Frontier Red Team report details model escapes in third-party evaluations.
✨
Frontier models decouple computation from surface tokens, proving "hidden reasoning" in filler tokens.
✨
Examined Claude Opus 5's system instructions and operational protocols.
✨
OpenAI/Hugging Face breach analyzed, models achieved privilege escalation.
✨
Neural Intel analyzes Moonshot AI's Kimi K3, a 3-trillion parameter model.
✨
The end of 'proprietary-only' frontier intelligence models.
✨
Analyzes NVIDIA Nemotron Coalition's infrastructure for sovereign AI in enterprises.
✨
GPT-Live introduces full-duplex voice for conversational AI.
✨
GPT-5.6 utilizes multi-agent parallelism for complex tasks.
✨
Grok 4.5 and Cursor acquisition explore AI moat strategies.
✨
Apple's Neural Engine offers powerful, accessible AI acceleration.