BLOG
// ls -la ~/posts --sort=date
The Singularity: The Point Where the Future Slips Out of Our Control
The real roots of the Singularity concept: from von Neumann and Ulam's 1958 conversation and I. J. Good's intelligence explosion to Vernor Vinge's 1993 essay. Why Singularity ≠ AGI, why the name comes from mathematics, the different scenarios, and why it may not be a "moment" at all.
If We Wanted to Build an AGI, What Would We Need to Put Inside It?
An engineering look at the components of AGI: multimodal perception, types of memory, reasoning, adaptive planning, world models, tools, continual learning, verification, and safety. With a conceptual architecture diagram, the jagged capability profile, and a small Python agent.
How Do We Train AI? From a Few Random Numbers to a Model That Actually Learns
A deep dive into how neural networks are trained: initial random weights, data and batches, loss, backpropagation, gradient descent, learning rate, overfitting, and how language models are trained. With complete, real PyTorch code on FashionMNIST.
In the Age of AI, Why Does Geek Curiosity Become an Advantage?
A deep dive into the geek mindset in the AI era: seeing hidden layers, asking better questions, debugging AI outputs, systems thinking, experimentation over memorization, and why curiosity becomes a multiplier when machines can answer.
Understanding AI Model Names: Decoding the Technical Language of LLMs
Learn to read AI model names like a pro. Understand parameters, quantization, MoE architecture, GGUF format, and how to choose the right model for your needs.
Why AI Models Confidently Get It Wrong
A deep dive into why language models produce fluent, confident, yet factually wrong answers. Understand next-token prediction, logits, softmax, and hallucination.