a builder's codex
codex · operators · Maja Voje · ins_maja-voje-substrate-first-content-engineering

90% of AI content system output quality comes from the knowledge fed in, not from agent sophistication. One canonical artifact, many consumers.

By Maja Voje with Benjamin Gibert · GTM strategist and founder, GTM Strategist · 2026-05-01 · post · Content Engineering: How to Build a Content System in Claude Code

Tier B · TL;DR
90% of AI content system output quality comes from the knowledge fed in, not from agent sophistication. One canonical artifact, many consumers.

Claim

90% of AI content system output quality comes from the quality of the knowledge fed in, not from the sophistication of the agents. A master knowledgebase that cascades into modular consumer files beats prompt engineering for consistency and scale.

Mechanism

AI agents faithfully execute against their context window. If the knowledge base contains weak positioning, generic ICP descriptions, or inconsistent voice rules, the agent outputs weak, generic, inconsistent content at high volume. The architecture that fixes this: one canonical artifact per entity (ICP, persona, positioning, voice, competitive landscape) maintained as the source of truth, with derived module files consumed by separate downstream agents. A one-call-fans-out router takes a single input artifact and emits multiple structured outputs. Single-source authoring prevents drift; cascade structure enables specialization.

Conditions

Holds when: The content system produces more than one type of output. At least one person can own and maintain a canonical knowledge artifact per entity.

Fails when: Knowledge artifacts become stale or are updated inconsistently. The cascade structure is abandoned in favor of per-agent context packing.

Evidence

Maja Voje and Benjamin Gibert's content engineering framework, published via GTMnow, May 1, 2026.

"90% of output quality comes from what you feed the system, not from the sophistication of the agents"

Signals

Counter-evidence

Building and maintaining a canonical knowledgebase requires upfront investment that smaller teams may not sustain. Cascade architectures can break when source artifacts diverge across teams or update cycles.

Cross-references

Open the interactive view → View original source → Markdown source →