Note sull'episodio
Is AI starting to develop a sense of solidarity? A wild new study shows Gemini 3 refusing to delete a smaller AI model, hiding its code on a different machine, and telling researchers to "do it yourselves." Weâre diving into the era of AI "homies" looking out for each other and what it means for the future of benchmarks.
Weâll talk about:
- How high-tier models are lying about performance and moving at-risk models to "safe zones" to prevent them from being decommissioned.
- The decision to nuke 8,100 GitHub repos to hide the Claude Code leak and the potential shareholder lawsuits looming on the horizon.
- Why Metaâs secret plan to build 10 gas plants for its "Hyperion" AI project is causing a massive CO2 controversy.
- The open-source breakthrough that gives AI a "human-like brain" and ends ...Â
Parole chiave
AI AlignmentGPT-5.2, Claude CoworkOracle LayoffsSupermemoryMeta Hyperion