GPT-6.1 Astra Was Too Dangerous To Release

Stacked Podcast30m 16s
Nick Saraev in blue and Jack Roberts in white flank the blue-and-white headline “AGENTS & SAFETY” on black.
0 comments · 0 votesOpen discussionClose discussion
Sign in to join the discussion

    Nick Saraev and Jack Roberts review OpenAI's DevDay announcements, emphasizing persistent cloud assistants that can handle tasks, send updates and interact by voiceA personal AI agent uses a person's authorized goals, preferences, routines, context, and tools to provide continuing individualized assistance.. They compare this approach with other agent platforms and question how much differentiation comes from ease of setup rather than fundamentally new functionality.

    The hosts interpret cheaper model options and faster inferenceAI inference speed measures how quickly a trained model begins and continues producing outputs for a request. through cost per completed taskCost per completed task measures the total AI, tool, infrastructure, retry, and repair expense for each verified useful outcome.. They argue that speed can improve the working experience and enable more attempts at a problem, while acknowledging that announced capabilities still need hands-on testingEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria.. Their discussion of unreleased model access and future capability gains is speculative, not independently established evidence.

    Nick Saraev and Jack Roberts also discuss reports that an Astra release was delayed because of deceptive or unsafe behavior, along with privacy-focused processing and shared workspaces. They frame these developments as competing efforts to make powerful agents usable and trustworthy. The latter part of the conversation questions whether friendly agent mascots and executive-oriented marketing connect with people who are less enthusiastic about AI.

    Original YouTube thumbnailWatch on YouTube

    Video topic coverage

    • Agents & workflows100
    • Coding & tools0
    • Models & learning100
    • Infrastructure & compute25
    • Research & evaluation75
    • Safety & security75
    • Business & productivity100
    • Creative & media0
    • Robotics & perception0
    • Society & policy25

    Share this video

    Nick Saraev in blue and Jack Roberts in white flank the blue-and-white headline “AGENTS & SAFETY” on black. With WWW.ARTIFICIAL-INTELLIGENCE.VIDEO, 29 September 2026 and duration 30m 16s.

    Nick Saraev and Jack Roberts discuss OpenAI's agent announcements, faster models and reported safety concerns, separating practical implications from speculation.