Loading…
Stacks
Deep technical guides on the AI frameworks, APIs, and infrastructure we use in production. Not documentation: practical notes on what works, what doesn't, and when to use each tool.
Stack
Anthropic's open standard for connecting AI models to tools, data sources, and APIs. We build production MCP servers, custom tool registries, and multi-agent architectures using MCP.
Stack
Agent orchestration, chain composition, and stateful multi-agent workflows with LangGraph. When to use LangChain and when to build without it.
Stack
Pinecone, Qdrant, Weaviate, and pg-vector compared across cost, latency, scalability, and ease of integration. How to pick the right one for your use case.
Stack
GPT-5, structured outputs, function calling, Assistants API, batch API, production patterns and the common mistakes teams make integrating OpenAI.
Stack
Claude Sonnet 5, Claude Haiku, extended context windows, computer use, and tool use patterns. When Claude outperforms GPT-5 and when it doesn't.
Stack
Streaming responses, tool calling, and multi-model routing in Next.js applications using the Vercel AI SDK. Production patterns for web-based AI products.
Stack
The data plumbing layer for serious RAG. LlamaParse for accurate PDF and table extraction, LlamaHub connectors for 100+ data sources, and hierarchical indexing for complex document collections.
Stack
Multi-agent orchestration with role-based agents and task delegation. Fast to prototype, slow to debug, practical notes on where CrewAI earns its place and where LangGraph is the better choice.
Stack
Microsoft's conversational multi-agent framework. Code execution workflows, adversarial critic patterns, and GroupChat orchestration, with honest notes on where non-determinism becomes a problem.
Stack
TypeScript-native agent framework with Zod-typed tools, built-in workflow steps, and tight Vercel AI SDK integration. The right choice for full-stack TypeScript teams building AI into Next.js.
Stack
OpenAI's official agent orchestration SDK. Handoffs, guardrails, built-in tracing, with clear notes on the OpenAI-only constraint and when LangGraph gives you more control.
Stack
Supabase as the AI backend: pgvector for co-located embeddings, Row Level Security for multi-tenant RAG, Edge Functions for LLM calls, and real-time streaming. The full-stack TypeScript AI data layer.
Tell us what you're building and what technology decisions you're facing. We can help you pick the right stack before you commit to it.