Posts

Paper Clusterer: Turn Your Obsidian Vault into a Research Map The Problem: A Graveyard of Research Notes My Obsidian vault had become a mess. Hundreds of paper notes, lecture materials, and research ideas—scattered across folders, poorly connected, and increasingly difficult to navigate. Sound familiar? As researchers, we consume vast amounts of information. We take notes, highlight papers, jot down ideas. But without proper organization, this knowledge remains fragmented and underutilized. ...

How foundation audio models and Audio LLMs bridge the gap between acoustic similarity and perceptual relevance TL;DR Acoustic similarity ≠ perceptual relevance. Two tracks can share timbre and tempo yet serve opposite listening contexts. Embedding models measure sound; Audio LLMs reason about context. Foundation models (MusicFM, CLAP, MuQ-MuLan) provide fast vector retrieval. Audio LLMs (Qwen2-Audio, Music Flamingo, Qwen3-Omni, LTU) add compositional reasoning and explainability. The future of music recommendation is fast retrieval + slow reasoning. A hybrid pipeline retrieves candidates via embeddings, captions them offline, and reranks with an LLM judge — then feeds those judgments back as hard negatives to strengthen the embedding space. Introduction ...

[AI for all] What is Intelligence? 🤔 So what exactly is intelligence? Intelligence is what makes us “Homo sapiens” - literally wise humans, thinking beings. What makes us intellectual is the fact that we can expect that the result of our actions will lead us to achieve our goals. But what makes machines intelligent? The same thing - their actions should lead to achieving their goals. The question is that we set the goals, and machines learn to optimize the process of achieving these goals in the best way. And we must be absolutely sure that we set the goals correctly, otherwise we can over-optimize, especially if (when?) we invent machines that will be much smarter than us and will quite be able to optimize us too :) ...

Beijing Car Visa: The Capital’s Invisible Border The role of capital cities is crucial in many countries, but in nations with governance structures similar to China—where the Party plays a leading role and the political system closely resembles the Soviet model—the capital’s role becomes particularly significant. While Russians joke about the “rubber city” of Moscow and its hyper-centralization, moving to or even driving to Moscow isn’t particularly problematic. You can drive into Moscow with regional license plates without major issues. ...

Education in China: From Disappointment to Opportunity This might be a bit rambling, but here we go. The Initial Dream My biggest disappointment at that time was probably my university experience. One of my main motivations for coming to China to study was to continue developing in mechanical engineering—to see how Chinese engineers work and how factories operate. Reality check: I didn’t set foot in a Chinese factory until 2024… but I’m getting ahead of myself. ...

Thoughts on Learning Chinese Yes, I speak, write, and understand Chinese. On one hand—it sounds impressive. On the other—why bother, really? The Practical Truth For the most part, Chinese turned out to be useful more for personal development (and for talking with my wife—obviously :)). Is it frustrating that the language is barely needed at work? I think so. Though when it was needed, Chinese corporate culture always came as a bonus… I should probably write about that separately sometime. ...