Position: Publisher
Reports to: Reality. 😄
Primary responsibility: Publish information so that both humans and machines can recover its intended meaning.
The publisher remains responsible for creating accurate, useful human-facing content. But the role now extends beyond the document. The publisher must preserve enough explicit knowledge structure that machines can identify what is being described, distinguish facts from surrounding prose, connect related entities, determine where assertions came from, understand when and where they apply, and resolve information needs without unnecessarily reconstructing knowledge the publisher already possesses.
And then I’d give the job actual responsibilities:
- Establish identity. Make entities uniquely identifiable and distinguishable from similar entities.
- Publish explicit facts. Preserve important assertions as recoverable knowledge rather than burying everything inside narrative.
- Preserve provenance. Record where facts came from, who published them, whether they were observed or derived, and which sources support them.
- Represent relationships. Make important connections among entities explicit rather than expecting machines to infer them.
- Define applicability. State when, where, for whom, and under what conditions knowledge is true.
- Preserve time. Treat versions, effective periods, supersession, and changing facts as part of meaning.
- Define terminology. Maintain stable definitions for domain language so facts and relationships are interpreted consistently.
- Design for resolution. Study information needs and publish the knowledge structures required to resolve them correctly.
- Maintain cross-surface consistency. When knowledge appears across websites, datasets, video, podcasts, documents, or other surfaces, preserve the same underlying meaning.
- Validate before publication. Test the publisher-controlled representation for completeness, consistency, identity, provenance, and structural errors.
- Observe after publication. Measure presence, citation, attribution, factual fidelity, relationship fidelity, applicability, provenance, and resolution in machine outputs.
- Improve the representation. Diagnose failures from observable behavior and correct what the publisher controls.
And then comes the part that makes this more than a checklist.
What is not in the job description
The publisher is not responsible for:
Training Google’s memory.
Conditioning ChatGPT.
Manipulating retrieval weights.
Engineering proprietary model confidence.
Forcing citation.
Creating separate “truths” for Gemini, Claude, ChatGPT, Perplexity, or whoever comes next.
Those things cross the publisher/consumer boundary.
The publisher’s responsibility ends with the integrity and quality of the reference it publishes.
After that:
Publish → Observe → Compare → Diagnose → Improve.
And I think there’s a killer closing formulation hiding here:
Old job description
Publish the document.
Make it discoverable.
Let the human resolve the meaning.
New job description
Publish the document.
Publish the knowledge behind it.
Make both discoverable.
Make the knowledge resolvable.
Observe what machines reflect.
Protect the meaning.