>_ dark-factoryGuides

Guides

Written by hand. New guides as they ship.

Practical pieces on building AI agent systems you own. Definitions, honest comparisons, and lessons pulled from the same running factory the build log documents. No roundups, no filler.

2026-09-02

Best AI Agent Courses in 2026: An Honest Guide

We sell one of these courses, so instead of a fake single ranking we compared by outcome. Every price and syllabus page-checked, gaps stated as what each page describes.

2026-09-02

OpenCode vs Claude Code for Building Agent Systems

We teach both. Open source and model freedom on one side, managed automation on the other, and the subscription fact most comparisons get wrong.

2026-09-02

ChatGPT Remembers You Now. Who Owns That Memory?

ChatGPT's memory is genuinely good in 2026. The harder questions, from OpenAI's own docs: can you see all of it, and can you take it with you?

2026-09-02

What Hardware You Actually Need to Run AI Agents

The agent layer needs 4 GB of RAM and no GPU. Local model inference is the only part with real hardware demands, and it is optional.

2026-09-02

Telegram AI Assistant: Run Your Own Agents From Your Phone

Not another chatbot in a chat app: Telegram as the free remote control for agents running on your own machine, from the docs, not from hype.

2026-09-02

What Is MCP? Explained for People Who Don't Code

The plug standard behind modern AI assistants, in household terms: what it is, who adopted it, and what it lets a non-coder actually do.

2026-09-01

Claude Code vs Cursor for Building Agent Systems

Every comparison asks which writes code faster. Ours asks which becomes a system you keep: whose cloud your agents run in, and whether they know you.

2026-09-01

The Rebuild: What Survived the February Wipe

A configuration reset took the whole workspace. What survived, what the recovery taught, and the rules the second system runs on.

2026-09-01

What Running AI Agents 24/7 Actually Costs

Fifty-four days of our own meter: 330 tasks, 3.37 million output tokens, about $265 at API-equivalent rates. Half the days cost nothing. Shipping days cost the most.

2026-09-01

Spec-Driven Development Tools in 2026: An Honest Field Guide

Spec Kit, Kiro, OpenSpec, BMAD, Cursor Plan Mode, and SpecWeaver, compared on the axis that separates them: who verifies the work, an agent's judgment or a machine check.

2026-09-01

Multi-Agent Orchestration Without LangGraph or CrewAI

The frameworks are good. This is the other architecture: independent CLI agents, a task board, and shared memory, with no framework code to write or maintain.

2026-09-01

Run AI Agents Locally: Models, Hardware, and the Token Bill

What Ollama and LM Studio actually do, which model sizes hold up in agent loops, honest hardware ballparks, and when local is not worth it.

2026-09-01

Personal AI Assistant with Memory: Buy One or Build One?

Every big assistant now remembers you, and that memory lives in their cloud. When built-in memory is enough, and when owning the files is the point.

2026-08-31

Claude Code for People Who Don't Code

You don't need a terminal in 2026. What the desktop app actually requires, what non-coders use it for, and how the tool becomes an assistant you own.

2026-08-31

GitHub Spec Kit vs SpecWeaver: Judged Specs, Checked Specs

Both are spec-driven development tools. One structures the workflow and lets the agent judge the result. The other makes the checks deterministic. They meet the spec at different ends.

2026-08-31

Claude Code 101 Is Free. Here's What It Doesn't Cover.

Anthropic's free courses teach the tool well. What no free course teaches is the persistent system you build with it: memory, identity, and a factory that runs while you are elsewhere.

2026-08-31

Hugging Face AI Agents Course: What You're Left With

An honest read on the free course: what its four units teach, what remains when it ends, and when a system you keep fits better than a certificate.

2026-08-31

What Is a Dark Factory?

The manufacturing term, where it comes from, and what it means applied to AI agents: a system that keeps working while you are not watching, on hardware you control.

Want the thinking before the guides? The free primer covers why chat has a ceiling and how assigning work to agents differs from asking questions. Ready to build? Start with Level 1.