Editorial Search Archive

Kimi K3 hit 270 HN points by running a 2.8T-parameter MoE model at 1 token/s on a MacBook Pro, streaming weights from four SSDs. This deep dive unpacks the SSD-streaming architecture and expert-rank-aware caching.

Fast-Agent is a zero-config, drop-in framework for composing MCP-enabled agent workflows in minutes. This guide builds a LangGraph pipeline that auto-discovers tools, negotiates schemas across servers, and routes tasks via a dynamic planner.

Muse Glimmer 30B — the 1,209-HN-point open-weight model engineered for always-on local agent inference. This guide builds a LangGraph workflow that runs entirely on commodity hardware with sub-500ms first-token latency.

Sim Studio hit 196 HN points as the Figma-like visual canvas for multi-agent workflow orchestration. This guide builds a LangGraph pipeline that connects MCP servers, LLM nodes, and routing logic on a drag-and-drop surface.

The Apple Health MCP Server hit 199 HN points by giving AI agents programmatic access to on-device health data. This guide builds a FastMCP server that reads HealthKit metrics, computes trend analysis, and surfaces live biometric streams.

A 570-HN-point MCP server proved you can cut Claude Code context consumption by 98% without losing accuracy. This guide builds a token-minifying proxy that compresses file trees, deduplicates log tails, and trims verbose tool outputs before they enter the context window.

A deep-dive analysis of speculative decoding — the LLM compiler optimization technique that uses a small draft model to propose tokens that a target model validates, cutting inference latency by 60% on production workloads without sacrificing output quality.

A production Stripe MCP server that exposes billing operations (customer management, subscription lifecycle, invoice handling, payment retry, refund processing) as AI-agent-callable tools — cutting billing operations overhead by 95% for SaaS platforms.

A production Redis Enterprise MCP server built with FastMCP 4.0 that provides distributed caching, session state management, pub/sub event channels, and vector similarity search for AI agents — cutting LLM response latency by 68% with semantic result caching.

A production GitHub MCP server that exposes repository operations (issue triage, PR review, CI/CD triggering, code search) as AI-agent-callable tools — triaging 93% of issues without human touch and cutting PR-to-merge cycle from 18 hours to 4.2 hours.

A production multi-modal document processing workflow that ingests PDFs, scans, and images — extracts text via OCR, classifies documents by type, generates embeddings, stores them in Qdrant, and enables semantic search — achieving 98.7% extraction accuracy and 95ms average query latency.

A production self-healing Kubernetes agent workflow that detects pod crashes, runs diagnostics, executes recovery strategies, and escalates to on-call engineers — cutting Mean Time To Recovery from 28 minutes to 3.1 minutes (89% reduction).