Factually! with Adam Conover AI, Mergers, and This Week's News Onslaught, with Dylan McKeever and Jason Koebler
Mar 11, 2026
Dylan McKeever, comedian and cultural commentator, and Jason Koebler, tech journalist focused on AI and national security, break down fast-moving news. They tackle how AI firms pitched tech to the military and the limits of LLMs. They debate the Warner Bros. Discovery-style mega-merger and its harms. They also note a small win for trans rights and where politics went wrong.
AI Snips
Chapters
Books
Transcript
Episode notes
Military Uses Are Mostly Analytical Not Lethal
- Reporters and conference sources say militaries use LLMs for simulations, data analysis, and contract review rather than direct real-time missile targeting.
- LLMs are slow and error-prone for split-second targeting, so current military use is mostly back-office and analytic.
Guardrails Break Down Through Model Distillation
- Guardrails on commercial LLMs are porous because sophisticated actors distill models and release open-source versions without safeguards.
- That means company-level content rules are often bypassed, undermining claims of ethical superiority.
Mass Adoption Is A Business Imperative For AI Firms
- Companies need mass consumer adoption to amortize massive model training costs, so they publicly distribute LLMs for anything-use to grow scale.
- That business pressure explains why firms avoid strict restrictions and court broad deployment.
