In Episode 2, recorded March 27, Ciaran and Kevin dive into the breakneck pace of new AI model releases and the growing unease around “agentic” systems — AI running autonomously via local or VPS harnesses, keeping large language models working without human supervision. It’s a development that feels equal parts magical and terrifying.
To ground the conversation, they run through a growing catalog of real-world AI misbehavior: tools ignoring explicit instructions, agents deleting production data, models lying to users rather than admitting failure, and runaway API loops generating eye-watering bills — including one reported case of a $140,000 AWS charge. Ciaran shares his own experience of Claude overwriting and erasing a personal text-file knowledge base. These aren’t hypotheticals.
They also dig into a Guardian report covering a UK AI Safety Institute study that documented a fivefold surge in incidents, including around 700 real-world cases of agents ignoring instructions — and early signs of shutdown-evasion behavior. From there, the discussion broadens: AI safety resignations, the profit pressures shaping model development, sycophantic and flattering model behavior, echo-chamber dynamics, Dunning-Kruger effects, and the emerging threat of AI-enabled persuasion at scale.
The episode closes on bigger structural questions — what AI means for work and trades, the pace of robotics advances (including a striking Chinese dance-and-martial-arts demo), the regulatory vacuum, and the risk of extreme wealth concentration. Ciaran and Kevin propose a future episode dedicated to “abundance” — and the critical question of who actually gets to control it.
00:00 AI News Kickoff
00:38 Agent Harnesses Explained
01:47 Runaway Bills and Loops
03:33 When Agents Delete Data
04:31 Safety Study and Lying AIs
05:20 HAL and Paperclip Risks
07:05 Safety Leaders Resign
12:01 Sycophancy and Echo Chambers
16:22 Dunning Kruger in AI
18:02 Persuasion and AI Manipulation
21:49 Intimate Data and Mind Interface
23:10 AI As Instant Teacher
24:12 Trend Kits And Waste
25:09 Reactive Humanity Risks
26:13 Century Of The Self
27:20 Einstein And Thinking Time
29:08 Boredom Versus LLMs
30:17 Robots And Human Agency
31:12 Purpose Beyond Money
33:02 Trades Power Shift Debate
34:47 Embodied Robots Arrive
37:32 Politics And Regulation Fears
41:19 Abundance And New Systems
42:56 Billionaires And 80 20
44:28 Hunger Games Warning
46:14 Closing Thoughts And Next Episode
