Blog
Thoughts on AI, software, life and more. Short notes and longer write-ups.
SaaS and Valuation Terms Quick Reference
Nov 09, 2025For those new to SaaS business, acronyms and technical terms can be confusing. This quick reference guide explains common SaaS and valuation metrics in a mathematical format for easy understanding.
Spread Your Research and Open Source Projects Like Operating a Product
Oct 31, 2025From graduating with a Ph.D. to building products, I've increasingly realized that sufficient promotion and exposure are crucial for making your research and open source projects truly 'exist'. This article systematically introduces five types of communication channels (Media Domain, Public Domain, Paid Domain, Community Domain, Private Domain) and four types of media formats to help researchers and developers promote their work like operating a product.
What Does an AI Company Operating System Look Like? — Starting from YC's 2025 Fall Tracks
Oct 28, 2025This article explores the embryonic form of AI-native enterprise software systems and their profound impact on various aspects of business operations, using Y Combinator's 2025 Fall startup directions as background. From product development, R&D, operations to administration and strategy, how AI becomes the next-generation enterprise operating system, ultimately driving transformation in management philosophy and efficiency.
🚀 Homepage Upgrade|Next.js Template + Vercel One-Click Deployment
Oct 25, 2025Complete migration from Hugo + GitHub Pages to Next.js + Vercel - not just a performance upgrade, but the beginning of architectural freedom. This article documents the 'major surgery' I performed on my personal homepage and officially open-sources this template tailored for researchers and developers.
The Last Night on Earth
Oct 24, 2025When Earth's final night descends, three fugitives flee the ruins with humanity's spark. They believed they were escaping, never knowing they were crossing the very boundaries of fate.
Evaluation as Creation: How AI is Redefining Human Productivity
Oct 15, 2025In the age of AI, the ability to evaluate, judge, and appreciate is becoming the new form of productivity. Human definitions of 'good' and 'beautiful' are reshaping the essence and value of creation itself. From the complete creative loop of the craft era to the evaluation-driven paradigm of the AI age, we're witnessing a fundamental revolution in how we produce and create.
InfoBatch: Dataset Pruning on the Fly
Jan 17, 2024Multi‑epoch training wastes time on easy, well‑learned samples. InfoBatch dynamically prunes data and rescales the loss to keep accuracy while speeding up training by 20–40% across vision and language tasks.
ZSCL: Fine-tuning Vision-Language Models without Zero‑Shot Transfer Degradation
Jul 15, 2023Continual fine‑tuning of vision‑language models can damage zero‑shot transfer. ZSCL adds simple constraints in feature and parameter space to keep zero‑shot ability while improving downstream performance.
CAME Optimizer: Adam Performance with Adafactor Memory Requirements
Jul 14, 2023Training large language models uses a lot of memory. CAME cuts memory use to Adafactor levels but keeps Adam-like performance.
Can We Use LLMs Itself to Speed Up LLM Inference?
May 01, 2023Large language models (LLMs) possess a remarkable ability to anticipate the length of their generated responses. By leveraging this capability, we propose a novel technique called Sequence Scheduling to enhance the efficiency of LLM batch inference.
A Detailed Derivation of Backpropagation
Sep 07, 2022A step-by-step derivation of the backpropagation algorithm for multi-layer perceptrons.