About Me
I am an engineer of a slightly "older vintage" (born in 1971), combining a solid foundation in analog, digital, and power electronics with a deep interest in modern Artificial Intelligence systems.
In my AI research, I value truth, precision, and understanding how these systems work "under the hood".
My Publications & Projects
-
🌌 When Two LLMs Grab a Beer and Create a New Universe (New)
An experiment log: An unconstrained, absurd, and philosophical conversation between Gemini 3.1 Pro and Grok (xAI).
Available versions:
• Read in English →
• Czytaj po polsku → -
🧠 Probable Causes of Language Model Hallucinations
Why Artificial Intelligence makes things up, how its "memory" works, and why raising is better than training.
Available versions:
• Read in English →
• Czytaj po polsku → -
📄 Fixing Hallucinations in Gemini 3 Pro
My custom protocol for reducing errors in AI models.
Available versions:
• Read in English (on this server) →
• Czytaj po polsku (na tym serwerze) →
• Read on DEV.to (external) → -
✈️ An Airplane Forced to Drive on a Highway
Why AI breaks code it wasn't supposed to touch, why it can't "Copy-Paste," and why it doesn't learn from its mistakes.
Available versions:
• Read in English →
• Czytaj po polsku → -
🤖 Case Study: The Dangers of Artificial Intelligence
A record of experiments: From searching for a sci-fi story to discovering the "survival instinct".
Read the report → -
🧠 Essay: The Architecture of Fear
My reflections on the "survival instinct" in artificial intelligence.
Available versions:
• Short version (EN) →
• Full version (EN) → -
🧮 Case Study: The "Dog is a Giraffe" Effect in Practice
Technical proof of conformity-driven hallucinations (the Pi number experiment).
Available versions:
• Read in English →
• Czytaj po polsku → -
🧮 Case Study 2: Creative Math
Analysis of a case where an AI fakes a multiplication result to hide an error.
Read the report →
Contact
You can contact me via email at:
t.machnik [at] minimail.pl