Most of these articles were vibe-coded. Every one was edited.
The Article Crew is how we turn a year of client work into a public knowledge base. Three agents do the legwork. A senior engineer is the editor. No piece publishes without a real human reading it end-to-end and signing it.
- researcher.agentSources, with a Singapore angle
Pulls primary sources (IMDA, MAS, IDA, Anthropic, OpenAI), recent academic, and live wGrow project notes. Always asks: what's the SG-specific take?
- drafter.agentLong-form, voice-controlled
Drafts in our house voice (declarative, short sentences, numbers over adjectives). Trained on a corpus of approved past articles, not on the open internet.
- fact-checker.agentCitations, numbers, claims
Flags every numerical claim and tries to ground it in a primary source. If it can't find one, the claim is rewritten or removed.
- editor (human)Read every word, then ship
A senior engineer reviews, edits structure, kills sycophancy, and either signs or sends back. We will publish slow and right.
- · Articles where a number couldn't be sourced.
- · Pieces that quote a "study" without naming the study.
- · Pieces that don't include something we learned the hard way.
- · Anything that sounds like a generic AI-agency LinkedIn post. We are too old for that.