All Jottings

March 16, 2026 · 39 min read

Humanity's Ultimate Divide in the Age of AI

The central question of the AI age is not how humans become more like AI, but how humans remain unlike AI.

March 12, 2026 · 38 min read

Diffusion Policy + RL: An Underrated Insight

Why diffusion policies fine-tuned with reinforcement learning work better than expected, through the lens of structured exploration, on-manifold search, and the geometry behind DPPO.

September 24, 2025 · 21 min read

From Gradient Descent to Reinforcement Learning: A Framework for Finding Research Ideas

A way to think about research ideas through the dual lenses of gradient descent and reinforcement learning, from solving known problems to exploring new paradigms.

September 13, 2025 · 7 min read

World Models Musing 1: On the Origins of Thought and the Three Great Schools

World Models: On the Origins of Thought and the Three Great Schools The Primal Impulse of Intelligence When we speak of “intelligence,” our minds often leap to images of Einstein scribbling on a blackboard or Mozart composing in...

May 08, 2025 · 12 min read

My CS Learning and Workflow: Harnessing LLMs, Obsidian, and Zotero

In the fast-evolving field of Computer Science, continuous learning is the norm. Daily work and study expose us to information from diverse, fragmented sources: hundreds of pages of official documentation, concise technical blog posts, industry news scrolled on...