When a Single Agent Provably Cannot
An Elimination Criterion and a Cross-Domain Taxonomy of Coordination-Forcing Conditions · Hifzullah Celik · July 2026
Most reasons we give for “needing a team” — too slow, too much to remember, not enough skill — dissolve the moment you imagine one agent patient, capacious, and capable enough. This paper isolates what is left. Posit an idealized single agent with unbounded patience, unbounded memory, and full capability, and admit a condition as coordination-forcing only if that agent still cannot do the task while a coordinated set can. What survives is a small taxonomy — ten conditions in five families, each a way that the singleness of a single locus of control, not its weakness, is the wall — together with a sharp negative result: redundancy for unreliable components is degradation, not forcing.
How to trust it
- Check it yourself. Every condition is an elimination argument you can verify by inspection against its named, real-world exemplars — the claims rest on those arguments, not on the author’s authority.
- Re-derived twice. The taxonomy was assembled a second time over a separately gathered corpus that shares no cases or conclusions with the first, and reproduced the same conditions and the same negative result — a robustness check across two independently assembled bodies of cases.
- Permanently archived. Published open-access with a citable Zenodo DOI (10.5281/zenodo.21227736), under CC BY 4.0.
- Ultradian · Texas, USA. Independent research, published on its own surface.