Research library
Software engineering
Practical architecture and development decisions for software that people can use, maintain, and adapt.
58 articles · Page 4 of 5
Anthropic Accidentally Leaks Claude Code's Entire Source Code via npm
A misconfigured npm package exposed 512,000 lines of Claude Code's source—revealing system prompts, KAIROS daemon, and unreleased features.
Gemini 3 Flash: Google's Best Coding Model Isn't Its Most Powerful One
Gemini 3 Flash: 78% SWE-bench outperforms Gemini 3 Pro on coding. Fastest, cheapest model optimized for agentic coding and multimodal reasoning.
NVIDIA Nemotron 3 Super: The Open Model Built for Multi-Agent Systems
Nemotron 3 Super: 120B open model blending Mamba and Transformer, with 4x memory efficiency, 1M token context, and 7.5x throughput advantage.
Anthropic's Claude Code Review: AI-Generated Code Reviewed by AI
Multi-agent code review for Claude Code: filters false positives, ranks bugs by severity, automates GitHub PR feedback with structured inline comments.
Small MoE Models: How Sparse Routing Makes Efficient AI Possible
Small-scale Mixture of Experts: sparse routing lets 47B models match 70B dense equivalents. Mixtral, DeepSeek-MoE, Phi-MoE, efficiency math.
GGML and llama.cpp Join Hugging Face: What It Means for Local AI
GGML creator Georgi Gerganov joins Hugging Face to secure the future of local AI inference. A landmark open-source move.
Agentic AI Architecture Patterns
A guide to agentic AI patterns: ReAct loops, tool-use protocols, multi-step planning, memory, and multi-agent coordination in production.
Mixture of Experts: Sparse AI Architectures
MoE architectures explained: gating mechanisms, expert routing, load balancing, and why sparse models deliver frontier AI at fraction cost.
MiniMax M2.5: Frontier Coding for Less
Chinese startup MiniMax matches Claude Opus 4.6 on SWE-bench while charging roughly one-tenth the price per token.
Building Software That Survives the AI Wave
The SaaSpocalypse is real for some software, overblown for others. Here's what AI can't replicate and how to build for it.
GPT-5.3-Codex: OpenAI's Agentic Code Model
OpenAI releases GPT-5.3-Codex with 77.3% on Terminal-Bench 2.0, an agentic coding model partially used in its own creation.
Foundations of Transformer Reasoning
A technical deep-dive into transformer architectures, attention mechanisms, scaling laws, and emerging techniques for reliable AI reasoning.