https://metacognition.computer/podcast/Thinking better, with AI. AI-generated deep-dives (Gemini Notebook; synthetic hosts) into the sessions of Jérémie O. Lumbroso's Penn course Computationally Assisted Metacognition — grades and reward functions, cybernetic vs. epistemic language, harnesses, and the admin password to your own brain. New episodes follow the course, weekly-ish. https://metacognition.computeren-us© 2026 Jérémie O. LumbrosoThu, 08 Oct 2026 18:24:42 +0000spotify@lumbroso.computer (Jérémie O. Lumbroso, Ph.D.)spotify@lumbroso.computer (Jérémie O. Lumbroso, Ph.D.)Hugo — self-hosted, cam-hq ADR-0009Jérémie O. Lumbroso, Ph.D.Thinking better, with AIThinking better, with AI. AI-generated deep-dives (Gemini Notebook; synthetic hosts) into the sessions of Jérémie O. Lumbroso's Penn course Computationally Assisted Metacognition — grades and reward functions, cybernetic vs. epistemic language, harnesses, and the admin password to your own brain. New episodes follow the course, weekly-ish. https://metacognition.computerepisodicfalseJérémie O. Lumbroso, Ph.D.spotify@lumbroso.computerhttps://metacognition.computer/sessions/session-02/metacogniting-002-2026-09-03Thu, 03 Sep 2026 20:00:00 -0400You have a system prompt. Nobody wrote it down, parts of it were installed before you could object, and you are running it right now. This episode dives into the second session of Computationally Assisted Metacognition (CIS 7000, Penn, Fall 2026): what it means to read the invisible instructions you inherited — and rewrite the ones you never agreed to; why venting to an AI before drafting the hard message changes what the draft becomes (and what context windows have to do with short-term memory); a professor's epiphany that began, of all places, on a toilet; and why the moral philosophy underneath it all — asking what no one could reasonably reject — turns out to be the most practical tool in the kit. About this show: episodes are AI-generated conversations produced with Gemini Notebook from Jérémie O. Lumbroso's own course materials; the two hosts are synthetic voices, and he introduces each episode himself. Course: https://metacognition.computer Corrections & precision notes: this episode is lightly edited — about 36 seconds were removed across twelve cuts where the synthetic hosts hallucinated details: an invented toilet-repair scene (the real story involves sitting on a toilet, thinking, not fixing one), a made-up term ("token memory"), an inverted etymology, a framework misattribution ("generative ethics" is a classification used in the course; the framework endorsed within it is moral contractualism), and a streaming-availability claim past its expiry date. Cut boundaries were placed by the professor's own ear on the waveform; what remains is what survived audit. The hosts' account of his personal disclosures is accurate and shared with his intent — it is course material, not leaked context.You have a system prompt. Nobody wrote it down, parts of it were installed before you could object, and you are running it right now. This episode dives into the second session of Computationally Assisted Metacognition (CIS 7000, Penn, Fall 2026): what it means to read the invisible instructions you inherited — and rewrite the ones you never agreed to; why venting to an AI before drafting the hard message changes what the draft becomes (and what context windows have to do with short-term memory); a professor's epiphany that began, of all places, on a toilet; and why the moral philosophy underneath it all — asking what no one could reasonably reject — turns out to be the most practical tool in the kit. About this show: episodes are AI-generated conversations produced with Gemini Notebook from Jérémie O. Lumbroso's own course materials; the two hosts are synthetic voices, and he introduces each episode himself. Course: https://metacognition.computer Corrections & precision notes: this episode is lightly edited — about 36 seconds were removed across twelve cuts where the synthetic hosts hallucinated details: an invented toilet-repair scene (the real story involves sitting on a toilet, thinking, not fixing one), a made-up term ("token memory"), an inverted etymology, a framework misattribution ("generative ethics" is a classification used in the course; the framework endorsed within it is moral contractualism), and a streaming-availability claim past its expiry date. Cut boundaries were placed by the professor's own ear on the waveform; what remains is what survived audit. The hosts' account of his personal disclosures is accurate and shared with his intent — it is course material, not leaked context.23:532falseJérémie O. Lumbroso, Ph.D.You have a system prompt. Nobody wrote it down, parts of it were installed before you could object, and you are running it right now. This episode dives into the second session of Computationally Assisted Metacognition (CIS 7000, Penn, Fall 2026): what it means to read the invisible instructions you inherited — and rewrite the ones you never agreed to; why venting to an AI before drafting the hard message changes what the draft becomes (and what context windows have to do with short-term memory); a professor's epiphany that began, of all places, on a toilet; and why the moral philosophy underneath it all — asking what no one could reasonably reject — turns out to be the most practical tool in the kit. About this show: episodes are AI-generated conversations produced with Gemini Notebook from Jérémie O. Lumbroso's own course materials; the two hosts are synthetic voices, and he introduces each episode himself. Course: https://metacognition.computer Corrections & precision notes: this episode is lightly edited — about 36 seconds were removed across twelve cuts where the synthetic hosts hallucinated details: an invented toilet-repair scene (the real story involves sitting on a toilet, thinking, not fixing one), a made-up term ("token memory"), an inverted etymology, a framework misattribution ("generative ethics" is a classification used in the course; the framework endorsed within it is moral contractualism), and a streaming-availability claim past its expiry date. Cut boundaries were placed by the professor's own ear on the waveform; what remains is what survived audit. The hosts' account of his personal disclosures is accurate and shared with his intent — it is course material, not leaked context.https://metacognition.computer/sessions/session-01/metacogniting-001-2026-08-27Thu, 27 Aug 2026 20:00:00 -0400What if the reason you feel overwhelmed by AI headlines isn't the headlines — it's that nobody ever handed you the admin password to your own thinking? This episode dives into the first session of Computationally Assisted Metacognition (CIS 7000, Penn, Fall 2026): why a professor gave every student an A on day one, what grades do to your reward function, the difference between using language to steer people and using it to understand the world — and why one professor sat in a parking lot for 45 minutes deciding about french fries, and what happened when he finally took the question seriously. About this show: episodes are AI-generated conversations produced with Gemini Notebook from Jérémie O. Lumbroso's own course materials; the two hosts are synthetic voices, and he introduces each episode himself. Course: https://metacognition.computerWhat if the reason you feel overwhelmed by AI headlines isn't the headlines — it's that nobody ever handed you the admin password to your own thinking? This episode dives into the first session of Computationally Assisted Metacognition (CIS 7000, Penn, Fall 2026): why a professor gave every student an A on day one, what grades do to your reward function, the difference between using language to steer people and using it to understand the world — and why one professor sat in a parking lot for 45 minutes deciding about french fries, and what happened when he finally took the question seriously. About this show: episodes are AI-generated conversations produced with Gemini Notebook from Jérémie O. Lumbroso's own course materials; the two hosts are synthetic voices, and he introduces each episode himself. Course: https://metacognition.computer22:141falseJérémie O. Lumbroso, Ph.D.What if the reason you feel overwhelmed by AI headlines isn't the headlines — it's that nobody ever handed you the admin password to your own thinking? This episode dives into the first session of Computationally Assisted Metacognition (CIS 7000, Penn, Fall 2026): why a professor gave every student an A on day one, what grades do to your reward function, the difference between using language to steer people and using it to understand the world — and why one professor sat in a parking lot for 45 minutes deciding about french fries, and what happened when he finally took the question seriously. About this show: episodes are AI-generated conversations produced with Gemini Notebook from Jérémie O. Lumbroso's own course materials; the two hosts are synthetic voices, and he introduces each episode himself. Course: https://metacognition.computer