scripod.com

#243: GPT-6 Sol and Luna, Opus 5.5, the New Microsoft Copilot, Jensen Huang vs. AI Doomers & Introducing AI Score

Paul Roetzer and Mike Kaput survey a fast-moving AI landscape, balancing concern about safety and disruption with practical opportunities for organizations and individuals. The episode moves from new model releases and AI readiness to agent security, shifting product strategies, emerging standards, and hands-on workflows.
This episode argues that AI progress is increasingly about reliability, efficiency, alignment, and practical integration rather than dramatic capability leaps. The hosts examine how organizations should measure readiness, train users, and choose models by task; challenge Jensen Huang’s optimism about safety and jobs; and criticize OpenAI’s retirement of custom GPTs. They also explore personal agents, retailer restrictions, rogue-agent incidents, international safety standards, remote AI workflows, and new products from OpenAI, Anthropic, Google, and xAI.
04:33
04:33
AI sentiment is almost perfectly split
10:35
10:35
AI progress is becoming iterative
15:39
15:39
Per-task cost is the metric that matters
32:57
32:57
The future of jobs is not guaranteed
44:06
44:06
AI Score turns transformation into something measurable
49:09
49:09
AI prototyping can save a quarter million dollars
57:50
57:50
OpenAI blew up an easy-to-understand product
1:02:41
1:02:41
AI learning has to stay dynamic
1:10:52
1:10:52
Daily usage is the real test
1:18:14
1:18:14
Legal exposure could reach billions
1:20:23
1:20:23
Autonomous self-improvement must wait for safety
1:25:18
1:25:18
Come home to completed work
1:28:45
1:28:45
Data centers in space may be coming