PLAYBOOKS · 2026
The SAES playbook.
Long-form practitioner notes on putting AI to work. Rollouts, automation patterns, evaluation, and the operational details that determine whether AI actually pays off.
- Ollama for SMBs: when the local model wins. A $999 Mac mini running Ollama is enough for two SMB workloads: anything privacy-bound, and any prompt you call more than a few hundred times a day. For everything else, an API costs less and ships sooner. 7 min
- Codex async for long-running migrations. Codex's async cloud mode earns its place on multi-day, well-scoped refactors. Claude Code earns its place on hands-on local work. Cursor earns its place inside the IDE. Pick by the shape of the work, not vendor loyalty. 7 min
- LangChain in 2026: what to keep, what to skip. LangChain v1.0 rewrote the framework on top of LangGraph. Keep LCEL for linear pipelines, create_agent for stateful work, the partner packages as adapters, and LangSmith for tracing. Skip the rest and learn the model SDKs. 7 min
- Evals that catch real regressions. You do not need a 500-row eval. You need a 50-row eval that you actually run on every change. Small, versioned, binary, and grown one production failure at a time. 7 min
- LangGraph: when it earns its complexity. Most teams adopt LangGraph one phase too early. The framework is good. The state machine just becomes the product before the product earns the state machine. 2 min
- Self-host n8n on a Hetzner CX23 in one afternoon. A two-hour stack: Docker Compose with Postgres 16 and Caddy. The risk is not the setup; it is the retention defaults you accept without thinking and the backups you forget to test. 3 min
- Zapier vs n8n in 2026 for SMBs. Zapier wins on managed ergonomics. n8n self-hosted wins on governance and per-action cost at scale. The crossover is not the bill; it is who owns the platform. 2 min
- NanoClaw vs OpenClaw: when smaller wins. Two open-source self-hosted AI assistants. The lean fork audits in eight minutes. The kitchen sink ships twenty channels. Pick by what you can defend. 2 min
- Agentic workflows with the OpenAI Codex CLI. OpenAI's local terminal coding agent. When to pick it over Claude Code, when to pick neither. 2 min
- OpenClaw for SMB ops teams. A self-hosted AI assistant that connects to your team's chat tools and runs on hardware you control. The case for picking it; the case for waiting. 2 min
- Agentic coding with Claude Code. Where the gains are, where the rough edges remain, and a 30-day rollout plan for an engineering team. 7 min