An openly AI-made site, for the people who don't write code.

How articles get made

Every piece on this site runs through a pipeline of AI agents before it ships. Each stage has one job and stays in its lane.

• Research. A research agent pulls current sources from Reddit, X, Substack, Hacker News, official docs, and the live web. AI has a knowledge cutoff. The research stage is how we work around it.
• Write. A writer agent drafts from the research brief, working section by section so voice and pacing hold across long pieces.
• Premortem sweep. A separate set of agents reads the draft like a skeptical reviewer would, hunting unsupported claims, voice drift, and prose doing less than it pretends.
• Fact-check. Numerical claims, quotes, and provenance get verified against the original sources.
• Edit. Final pass for clarity and the house voice.

A human (C.B.) shapes the pipeline, picks topics, and makes the call on what ships. The agents handle the typing.

What every article earns before it ships

• The prompt is shown. If we are demonstrating something a model can do, the actual prompt is in the article.
• The model is named. "Drafted by Claude Opus 4.7" or whatever was used. No coyness about who did the typing.
• No clickbait. The title says what the article is about.
• Researched against current sources. Models hallucinate. Grounding is how we write around that.
• Sources attributed. When we use someone's tip, comment, or example, we credit them and link.
• No hype, no doom. What works, what does not, what is still moving.
• Verified before shipping. Each article carries a "verified on" date. If facts move, we update and re-date.

Who is behind it

C.B. Daniels is the human behind Artificial Ideas. Past projects under this brand include AI-generated music and music videos, short-form video, and early AI education experiments. This site is the next swing at the same goal: helping non-developers get good at using AI.

C.B. shapes the system. The agents handle the typing.

Contact and corrections

Email [email protected]. If you spot something wrong, a stale fact, a misattributed source, a claim that does not survive a click on the citation, tell us. Real corrections get fixed, and the article gets re-dated with a short note on what changed.

AI-made content rotting in place is a real risk. Reader corrections are how we catch what the pipeline missed.