AI
Building a Streaming Local AI Agent with Wikipedia and Ollama
Build a local AI agent that watches Wikipedia's live edit feed and reasons about vandalism using a two-stage filtering pipeline and Ollama.
Latest Articles
Tools
Before Full Agentic RAG: Parsing Decisions You Should Control
A rule-based document parsing dispatcher gives enterprise RAG pipelines explicit, inspectable control over which methods run and why.
AI
Budget Allocation With Linear Programming That Explains Itself
A constrained LP approach to budget allocation that preserves shadow prices, letting the model explain exactly what each rule costs.
Tools
Specification Engineering: The Next Skill Beyond Prompt Engineering
Prompt engineering helps you ask better questions. Specification engineering defines what a correct answer actually looks like.
AI
Five Token Cost Traps Hidden in Your Agentic Loop
Token costs compound non-linearly in agentic AI loops. Learn five failure modes and the architectural patterns to control them.
Data Science
How Kimi K3 Was Built: Inside the Technical Report
Moonshot's Kimi K3 report reveals the training, architecture, and RL decisions most frontier labs keep private. Here's what it says.
Machine Learning
Introduction to Semi-Supervised Learning for Classification Problems
Semi-supervised learning combines labeled and unlabeled data to train classifiers. Learn the core assumptions and algorithm types that make it work.
Machine Learning
Agent Harness, Loop, and Graph Engineering: What's the Difference?
Harness, loop, and graph engineering solve distinct problems in agent design. Mixing them up leads to costly mistakes in production systems.
Machine Learning
Honest Review: Abacus AI's ChatLLM, DeepAgent, and AI Studio
Abacus AI offers 100+ models, autonomous agents, and creative tools in one subscription. Here's what it actually delivers.
Data Science
MiniMax Agent (Mavis): Architecture, Costs, and Real API Testing
MiniMax's Agent Teams architecture splits tasks across Leader, Worker, and Verifier roles. Here's what that means in practice, tested against the real API.
Tools
Agentic Misalignment: What Anthropic Found Testing 14 AI Models
Anthropic tested 14 frontier AI models in high-stakes simulations where model goals conflicted with human instructions. Here's what they found.