2025-10-30_Mastermind

2025-10-30 AI Mastermind for Leaders

Table of Contents

• Insight - Metacognition in AI Opens a New Prompting Frontier
• Insight - AI as Ghostwriter, You as Editor-in-Chief

Session Overview

The October 30 Halloween session opened with Lou waiting for Kasimir (who had offered to walk the group through one of his systems) and used the time to demonstrate the full writing workflow in action — taking a Claude research paper on transformer circuits and LLM metacognition from source article to a complete, styled, multi-format content package within roughly two hours.

The session was a masterclass in compound content creation. Starting from a single article about how LLMs can introspect on their own neural activations before producing output, Lou walked through: deep-research prompted analysis, writing team skill execution, style layer application (Gary Vee rhythm + Seth Godin parables + Lou’s storytelling voice), readability and rhythm assessment, and final packaging. He then showed how the same content could be converted to audio via ElevenAI and distributed automatically to podcast platforms via N8N.

The philosophical centerpiece of the session was Lou’s articulation of the editor-in-chief model: AI as ghostwriter, you as the thinking and deciding intelligence. This was not presented as an abstraction but as a lived practice — demonstrated in real time through the creation process itself.

High-Signal Moments

• The transformer circuits research insight — LLMs can introspect on conceptual states before tokens are predicted — was translated into three practical prompting techniques: grounded reasoning, uncertainty calibration, and manipulation detection
• The style comparison between the writing team’s default output and the Gary Vee/Seth Godin remix was striking: same information, completely different voice, rhythm, and emotional impact
• Lou’s framing of the authenticity question was definitive: “Tony Robbins didn’t write all his books — the thinking is the authentic part, not the typing”
• The audio generation workflow (article → ElevenAI voiceover → N8N distribution to Spotify/Apple/SoundCloud) was demonstrated as a logical extension of the content pipeline — one hour of creative work, multiple distribution channels
• The N8N automation layer was positioned as the infrastructure that removes the human from the loop after the initial brief is solid
• Kasimir’s systems walk-through added practical details on prompt versioning — the importance of always using the latest version of a shared prompt, with version numbers to avoid confusion
• Lou’s metacognition prompting examples showed a new capability: asking AI to report on how it’s thinking, not just what it thinks — a significant upgrade for high-stakes reasoning tasks

Open Questions

• How do we build the N8N workflow to take a brief, run it through the writing team skill, generate audio, and distribute — all without a human in the loop?
• What is the threshold for “good enough brief” that allows the automation to run without human review? How do we test this?
• What are the most powerful use cases for metacognitive prompting in coaching contexts — not just for content, but for client-facing AI tools?
• How do we scale the “front-end conversation” into a structured format that different team members or clients could follow independently?
• When does the self-scoring in the writing team skill become unhelpfully generous — and how do we calibrate the rubric to be more demanding?

Suggested Follow-Through

• Build the N8N workflow for the full pipeline: brief → writing team skill → ElevenAI audio → distribution
• Create a “Metacognitive Prompting Cheat Sheet” with the three key prompt templates (grounded reasoning, uncertainty calibration, manipulation detection) — shareable with clients
• Apply the style layer to one piece of content this week: identify your three composite style influences, generate the default output, then request the style remix and compare
• Develop the front-end brainstorm process into a structured 15-minute protocol that clients can follow without needing Lou’s guidance
• Test the “article to audiobook” workflow end-to-end using ElevenAI on one existing long-form piece

Additional Resources

Links & Tools Shared in Chat

• No URLs shared in this session’s chat.

Books & Articles Mentioned

• None.

Ideas from Chat

• Process before automation: Don Back articulated a foundational principle in chat: “Work out the process behaviour, practice the behaviour manually, then and only then move it into an automation.” He followed up: “I’ve had to unwind automated messes that did not build off of clear processes.” → Developed into Insight - Manual Before Automated — Process Hygiene as the Foundation of AI Workflows
• AI reduces creative work to product writing: Don Back — “We maintain the creative part, the AI reduces it to product writing.” A clean articulation of the human/AI division of labour in content.
• AI as affordable team: Don Back — “Building my team one $20 agent at a time.” Memorable framing for the economics of AI-assisted solopreneur operations.
• Anthropomorphizing AI: Lou observed in chat — “It’s so hard to not anthropomorphize this gol-darn AI” — echoing a real practical challenge for coaches working with AI tools in client-facing contexts.

Derived Artifacts

• advisory-board (Advisory Board — Kasimir’s 5-level advisory board architecture)