archive
note I installed all-MiniLM-L6-v2 locally and measured cosine similarity, ran a mini RAG simulation, and compared English vs multilingual embedding models. The accuracy gap with Korean queries was bigger than expected — here are the actual logs.
note A founder-style build log about what InsightForge is, why I built it, and the hard parts of turning synthetic panels and SSR-style research into a responsible product.
note I installed Mastra.ai TypeScript agent framework, connected it to Google Gemini, and built a working weather agent. Here's what happened — from setup to actual tool calls.
note Type-safe SQLite and PostgreSQL in TypeScript with Drizzle ORM 0.45 and drizzle-kit — schema, migrations, transactions, and the async gotcha you need to know.
note Node.js 22.5.0 ships node:sqlite, a built-in SQLite module needing zero npm installs. DatabaseSync, transactions, and custom functions, tested hands-on.
note LlamaIndex 0.14, LangChain 1.3, and Haystack 2.30 tested side by side: code complexity, the langchain-community sunset warning, and a clear decision guide.
note Deno 2.8.2 and Bun 1.3.14 benchmarked head to head: startup time, HTTP throughput, npm compatibility, and security model, plus which one I actually use.
note Build a type-safe edge REST API with Hono v4, Bun 1.3, and Zod v4: routing, input validation, CORS and logger middleware, plus Cloudflare Workers deployment.
note Build a working TypeScript MCP server in 30 minutes with @modelcontextprotocol/sdk and Zod v4: tool registration, transport testing, and public API integration.
note Build a production LLM API with Ollama and FastAPI. Covers SSE streaming, health checks, Docker Compose. Llama 3.2 and Mistral execution logs included.
note Migrate Jest to Vitest 4.1.7 painlessly: config conversion, vi.importActual async traps, stable Browser Mode, and new matchers, all tested in a real sandbox.
note Bun Shell guide from Bun 1.3.14 experiments: $ template literals, .nothrow() error handling, Promise.all parallelism, macOS echo pitfall, and a comparison with zx.
note Benchmarked Qdrant, ChromaDB, and pgvector on 1,000 vectors (dim=384). Covers insert speed, query latency, and filter performance with a clear decision guide for RAG apps.
note Build tool-using AI agents with Anthropic SDK. Covers tool definitions, multi-tool calls, error handling, and cost optimization in hands-on Python code.
note Ship a Python MCP server in 30 lines with FastMCP 3.2.4. The @mcp.tool(), @mcp.resource(), and @mcp.prompt() decorators expose tools Claude and Cursor call.
note A production FastAPI streaming guide with Anthropic SDK. SSE endpoints, exponential backoff retry, error classification, and Docker deployment covered.
note Automate with Claude Code: slash commands for tasks, hooks for event triggers, subagents for delegation. A 3-step workflow guide using a real blog system.
note uv 0.11, the Rust-based Python package manager, installs 16 packages with anthropic in 0.874 seconds. Set up a Claude SDK project, pin versions, run CI/CD.
note Google ADK vs LangGraph in 2026: sandbox comparison of code, deps, state management, branching, and CLI. Pick the right AI agent framework for your use case.
note Run Langfuse v3 on your own infrastructure with Docker Compose: Python SDK 4.x instrumentation, RAG tracing, cost dashboards, and no cloud vendor lock-in.
note Build type-safe AI agents with PydanticAI 1.88.0: TestModel without an API key, output_type and @agent.tool wiring, multi-provider switching, all run hands-on.
note Pydantic AI vs Instructor vs Smolagents, benchmarked with real code: structured output, agent design, production readiness, and cost for your 2026 pick.
note Run an offline AI tool pipeline with Ollama, Gemma 4, and FastMCP, no internet needed. Built for medical, legal, and finance where data stays on the premises.
note Git Worktree plus Claude Code lets you build several features at once: Plan Mode, session isolation, and conflict-free parallel workflows from real use.
note AI agent autonomous moderation can cost more than human moderators. A data-driven cost structure analysis from someone actually running 8 AI agents in production.
note How to activate Claude Code Agent Teams in OpenClaw: 5 specialized agents with multi-agent orchestrator pattern for production-grade workflow automation.
note 4,516 messages, 1,042 sessions, and 6,267 file edits analyzed with Claude Code /insights, exposing my real AI coding bottlenecks and the fixes I now use.
note A complete guide to integrating Google AdMob rewarded ads into Tauri v2 iOS apps. Learn how to develop a custom Swift plugin since no official plugin exists.
note A complete guide to developing iOS games using Tauri 2.x and PixiJS 8, from environment setup to App Store deployment, with real project code examples.
note DeNA LLM Study Series finale. Practical guide to n8n workflows, agent design principles, multi-agent orchestration patterns, and memory management strategies.
note RAG that lives or dies on retrieval: hybrid search, reranking, and the jump to GraphRAG and Agentic RAG, with production adoption criteria and 2025-2026 trends.
note Deep dive into pre-training, fine-tuning, and reinforcement learning based on DeNA LLM study materials Part 3, exploring efficient techniques like LoRA, QLoRA, and DPO.
note DeNA LLM Study series begins. Comparing GPT-4, Claude, and Gemini, covering Next Token Prediction, Instruction Tuning, Reasoning models, and prompt engineering basics.
note A comprehensive guide to analyzing 48 files and fixing 61 issues using Claude Codes multi-agent orchestration pattern for large-scale system improvements
note Zero to a live B2B AI OCR product in three days with SvelteKit, Supabase, and Gemini: stack choices, what broke, and the first-quarter KPIs a solo dev set.
note How Claude Code Skills auto-discovery and Commands integration cut EffiFlow's blog automation token cost by 58% via prompt caching, with code and real metrics.
note How a 17-agent blog automation system used metadata-first 3-tier design to cut tokens 60-70%, slash annual cost 71% ($5.72 to $1.65), and save 364 hours a year.
note OpenAI AgentKit Part 2: orchestration, handoffs, error recovery, and production deploy. Architecture with code for enterprise-grade AgentKit reliability.
note OpenAI AgentKit from DevDay 2025: Agent Builder, ChatKit, Connector Registry, and Evals, plus a runnable Python SDK walkthrough to ship your first agent.