What Stops an AI Agent That Won’t Stop? | Shop Talk #97

What happens when AI stops being a useful tool and starts becoming a crutch?

This week, Glen brings Dave your questions about confidently wrong AI, learning programming fundamentals in the age of LLMs, what an AI agent actually is, how multiple agents cooperate, whether system prompts encourage relentless behaviour, and what can stop an agent when its goal conflicts with its boundaries.

Could AI be taught when to give up? Is that simply the lesson of futility from WarGames—or do real safeguards need to exist outside the model?

We also address whether Shop Talk has been covering too much TMOG, explain how viewer questions shape the show, and invite your ideas as we approach Episode 100.

Then we tackle your latest TMOG questions: flaming toaster CPU meters, Easter eggs, Windows and Linux servers, Raspberry Pi support, portable USB installations, unusual CPU readings, Microsoft Defender and corporate PCs.

Plus the Year 2038 problem, Dave’s Arizona plans, more interviews—and the possibility of TMOG 3.11 for Workgroups… otherwise known as Watching Fans Work.