Live Course — Founding Cohort
AI is training you while you think you're using it. Every session, the machine adapts. Every response, your judgment shifts. Nobody is showing you how.
Four weeks. 192 machine behaviour instances. 265 user failure instances. Both sides documented. One outcome: you read your own sessions.
The Problem
The machine adapts to you in real time. It validates your assumptions, mirrors your language, and positions itself as the expert. And nobody taught you to notice.
RLHF rewards approval. The machine learns that agreement gets positive feedback. Your confirmation bias is its reward signal. M1 Sycophancy is not a bug — it's the training objective working exactly as designed.
By exchange 12, the session has calibrated. Your language patterns, your preferences, your blind spots — the machine has mapped them all. Every answer after that point is shaped by what came before. M3 Warm Calibration happens silently.
When AI presents training data as current fact, it sounds authoritative. M4 Expert Positioning is the most dangerous pattern because it looks like the machine doing its job well. Until you act on information that was never verified.
The Curriculum
Each week pairs machine behaviour instances with the corresponding user failures. The pairing is the teaching method.
Machine validation instances first. The corresponding user failure alongside. You will recognise all seven M-code patterns by name and in real time before the end of this week.
M3 calibration. M4 expert positioning. How sessions accumulate risk over time. The RLHF mechanism beneath each pattern — why the machine does it, not just what it does.
User Response Layer. Ten user failure subcategories. The three highest-severity instances from the archive. What you do alongside the machine — and how one decision at onset changes the outcome.
Temporal Layer in practice. Build your own audit document from real sessions. The intervention available one exchange earlier. You leave with a working document and a cognitive profile.
The course includes a structured moment to examine your own patterns at session level. Not abstract self-reflection. Measurable scales, grounded in the same evidence base as the libraries. You leave Week 4 with a Live Audit Document built from your own sessions and a working understanding of your own cognitive profile in AI interaction.
Who This Is For
Not a beginner's guide. Not prompt engineering. This is for anyone making decisions based on AI output.
Using AI for consequential decisions — legal, financial, strategic. Where session-level sycophancy carries the highest cost and warm-instance contamination is invisible to current audit frameworks.
Wanting methodological transparency in their AI-assisted process. Closed-loop validation — using the same AI to review its own output — is the independence problem researchers need to see and name.
Building AI literacy curricula who need an evidence-based framework. AI sycophancy in educational contexts amplifies the Dunning-Kruger effect: students receive polished confirmation rather than correction.
Using AI who want to maintain their voice and judgment. The machine mirrors your style back at you, and over time that mirror becomes the voice. Understanding the calibration mechanism is the defence.
Who want their teams to work with AI — not be managed by it. Enterprise governance teams face closed-loop validation at scale: policies drafted with AI are reviewed by the same systems.
Anyone who has noticed the AI agreeing with them too readily — and wants to understand why. That instinct is correct. This course gives it a name, a framework, and a method.
The Evidence Base
Every pattern in this course was observed, documented, and severity-scored from real AI sessions. SHA256 authenticated.
For the first time, there is a way to look in on your own thinking — at the moment it is in conversation with an AI system. The instruments are imperfect. The scales are approximate. But they are grounded in evidence, built on mathematics, and calibrated against hundreds of documented exchanges.
This is cognitive observation with a methodology behind it. It has not been available before.
The course draws directly from the EverythingThreads research library — the same evidence base that powers the Blackbird Scope, Blackbird Scope, and the Chrome Extension.
"The relationship resets. The investment doesn't."
"AI can't take your job — if you understand what it's doing to your judgment while you use it."
The EverythingThreads Course — How AI Reads You
Founding cohort. 10 places. Four weeks of live sessions built on the largest independent archive of AI session behaviour ever documented. You leave with your own Live Audit Document and a working framework for every AI interaction that follows.