Archives

Worked Python solutions to the LeetCode Top Interview 150, organized by topic with a short approach for each problem.

A ground-up tour of how neural networks learned to see, then to see-and-read, and finally to imagine. From AlexNet and CNNs, through CLIP and the vision-language models behind GPT-4V, to world models like Dreamer, V-JEPA 2, and LeWorldModel — with architectures, math, and benchmark numbers along the way.

A deep dive into the Kimi team's Attention Residuals (AttnRes) — replacing the fixed-weight residual connection with learned softmax attention over depth. Covers the time–depth duality, Full vs Block AttnRes, the structured-matrix view that unifies prior residual variants, the pipeline-parallel infra that makes it practical, and the scaling-law and 48B-MoE results.

An end-to-end walkthrough of Group Relative Policy Optimization (GRPO) and Decoupled Clip and Dynamic sAmpling Policy Optimization (DAPO) — the two RL algorithms that drive open reasoning models in 2025–2026. Full math, every design choice motivated, and a head-to-head comparison.

A complete walkthrough of Group Relative Policy Optimization (GRPO) and Group Sequence Policy Optimization (GSPO) — the policy-gradient algorithms behind DeepSeek-R1 and Qwen3. Full math, the failure mode that motivated GSPO, the MoE story, and a side-by-side comparison.

An end-to-end derivation of Group Relative Policy Optimization (GRPO) from DeepSeekMath and the Dr.GRPO correction from Liu et al. Covers the full objective, the gradient, the two biases (length and question difficulty), the unbiased fix, and the practical recipe behind R1-Zero–style training.

A structured walkthrough of Sasha Rush's Training Composer 2 workshop: why Cursor chose Kimi K2.5, how continued pretraining and long-horizon RL fit together, what CursorBench measures, and where Composer is headed.

Leetcode grinding.

A side-by-side dive into Xiaomi MiMo's hybrid sliding-window/global attention and DeepSeek's Multi-head Latent Attention. The two answer the same question — how to make attention affordable at long context — with very different bets, and those bets shape everything from training infra to KV cache size.

A walkthrough of two ideas behind Kimi K2.5: how joint text–vision pre-training and RL make each modality help the other, and how Agent Swarm replaces sequential tool use with a learned parallel orchestrator.

A deep dive into DeepSeek's two sparse attention designs — Native Sparse Attention (NSA) and DeepSeek Sparse Attention (DSA) — covering the math, the hardware story, and why DSA in V3.2 looks so different from NSA.

AstroPaper v6: a from-scratch rewrite on Astro v6, Tailwind v4, and a new config system.

AstroPaper v5: keep the clean look, updates under the hood.

Learn how to add LaTeX equations in Astro blog posts using Markdown, KaTeX, and remark/rehype plugins.

Comment function on a static blog hosted on GitHub Pages with Giscus.

AstroPaper v4: ensuring a smoother and more feature-rich blogging experience.

How to use Git Hooks to set your Created and Modified Dates on AstroPaper

AstroPaper Version 3: Elevating Your Web Experience with Astro v3 and Seamless View Transitions

How to update project dependencies and AstroPaper template.

AstroPaper with the enhancements of Astro v2. Type-safe markdown contents, bug fixes and better dev experience etc.

New feature in AstroPaper v1.4.0, introducing dynamic OG image generation for blog posts.

Some of the well-crafted, updated predefined color schemes for AstroPaper.

How you can enable/disable light & dark mode; and customize color schemes of AstroPaper theme.

Some rules & recommendations for creating or adding new posts using AstroPaper theme.

How you can make AstroPaper theme absolutely yours.

EXAMPLE POST: About Tailwind Typography Plugin and how you can use it effectively.

EXAMPLE POST: Developing a terminal-like website using ReactJS, TypeScript and Styled-Components. Includes features like autocomplete, multiple themes, command hints etc.

EXAMPLE POST: My experience about developing my first portfolio website and a blog using NextJS and a headless CMS.