Project Origins

Originally conceived as "HITL-BOS" (Human-in-the-Loop Business Operating System). Abandoned due to phonetic problems in English.

"Directed Systems" was chosen because it maps to an existing concept: Sheridan and Verplank's levels of automation (1978). "Directed" describes configurations where AI systems hold significant decision-making authority while humans retain execution and oversight.

The initial version had several problems:

It claimed synthesis while proposing invention. Terms like "Taskifier," "Operational Framework," and "Three Powers" were presented as if derived from research. They were not. They were invented vocabulary dressed in academic language.

It made predictions without basis. "Perhaps five years. Perhaps fifteen." This was speculation, not analysis.

It used unnecessary jargon. "High-Fidelity Biological Interface" meant "the human." The term added complexity without clarity.

It claimed no originality while hinting at vocabulary ownership. That contradiction weakened credibility.

The second version tried to separate research from proposals. It improved the document, but still read too much like a hybrid between manifesto and literature review.

That pass clarified the research base, removed invented terminology, and made the work more defensible. It still had a problem of purpose.

The January public release reframed the document as a practical guide. The central question became: why are humans still necessary, and how should people prepare?

That version was stronger than the earlier drafts because it was:

• more grounded in research
• less inflated in tone
• more useful to a general reader

It was still limited. It remained closer to a guide than to a paper. It explained the shift, but it did not yet contribute a sharper analytical lens for describing the emerging configuration itself.

The March 2026 release for v0.2.0 moves again, but in a narrower and more deliberate way.

The document is no longer framed primarily as a guide to preparation. It is now a practitioner paper that tries to do one specific thing well: identify a recurring organizational pattern in AI adoption and describe its consequences without overstating novelty.

The main change is analytical. The current draft argues that many AI workflows are best understood as directed systems and proposes three properties for recognizing them:

• shifted initiative
• retained human consequence
• asymmetric learning

This framing is intended to add value beyond summary. It connects contemporary AI workflows to older automation concerns such as skill decay, substitution myths, and responsibility without pretending to offer a new scientific taxonomy.

From the first drafts to the current v0.2.0 release, the trajectory is:

1. remove invented language
2. remove unsupported prediction
3. increase empirical grounding
4. lower rhetorical temperature
5. add a bounded analytical contribution

The goal is not to sound academic. It is to become more exact.

A reasonable test is no longer "does this sound ambitious?" It is: does this add a frame that helps someone analyze real AI-mediated work more clearly than they could before?

If the answer is no, the paper should be cut back or rewritten.

If the answer is yes, then the document has earned its length.

March 2026

Petru Arakiss