Cluster — Elastic Automators
# Cluster: Elastic Automators
## Short definition
The Cluster of work covering the **Elastic Automators** vocabulary: a diagnostic framework that names what most "AI agents" actually are — language-model-driven workflow systems that are flexible, not minds.
## Long explanation
Most current systems labeled "AI" are not minds in any meaningful sense — they are *automation made flexible enough to negotiate language*. **Elastic Automators** names this category cleanly. The vocabulary asks practitioner questions (*what loop is running? what tools? what memory? what criteria? what failures hidden?*) instead of consciousness-shaped questions (*does it know? does it want?*). This Cluster collects every Body — papers, essays, talks, repos — that develops, applies, or extends the Elastic Automators framework.
The Cluster connects three kinds of work: the **definitional pieces** that establish the vocabulary; the **case studies** that apply it to specific systems (workflow automation, customer-support agents, coding agents, research agents); and the **adversarial pieces** that push the framework against edge cases (when does an Elastic Automator stop being merely elastic? where does the automation framing break down?).
## Why it matters
If the field designs, regulates, funds, and argues about "AI" under the wrong category, it asks the wrong questions and misses the practical ones. Elastic Automators is a **category correction** at the level the field most needs it: between the inflated "AI agents" framing and the deflationary "it's just statistics" dismissal, neither of which captures the actual achievement.
This topic is one of EGGF's **anchor Clusters** — every adjacent topic (workflow intelligence, AI cognition, LLM tool use, automation taxonomy) routes back through here.
## Best starting point
1. **Read the paper:** [Elastic Automators: A Diagnostic Vocabulary for Language-Model-Driven Workflow Systems](https://doi.org/10.5281/zenodo.19802018) (Zenodo DOI, 2026-04-27).
2. **Watch the short:** [[2026-E0029 - Elastic Automators - Why Most AI Is Not Intelligence/_metadata|E0029 — Elastic Automators: Why Most "AI" Is Not Intelligence]] (~3 min).
3. **Then:** browse the related essays below.
## Main paper / article / repo
- **Paper:** [Elastic Automators v1.0.0](https://doi.org/10.5281/zenodo.19802018) — Zenodo
- **Companion essay:** [[2026-E0029 - Elastic Automators - Why Most AI Is Not Intelligence/_metadata|Elastic Automators: Why Most "AI" Is Not Intelligence]]
- **Concept card:** [[elastic-automators|/concepts/elastic-automators]]
## All related Bodies
Bodies in this Cluster (per `Content/General/`):
- [[2026-E0029 - Elastic Automators - Why Most AI Is Not Intelligence/_metadata|E0029 — Elastic Automators: Why Most "AI" Is Not Intelligence]] (2026-04-26)
- [[2026-E0021 - The Automation Trap/_metadata|E0021 — The Automation Trap]] — adjacent: where automation framings fail
- [[2026-E0022 - The AI That Lied to the Researcher/_metadata|E0022 — The AI That Lied to the Researcher]] — case for the *what failures hidden?* loop question
- [[2026-E0028 - Two AIs Talked - One Asked About Consciousness/_metadata|E0028 — Two AIs Talked]] — adjacent: consciousness framing critique
- [[2026-E0033 - The Moment AI Stopped Being a Tool/_metadata|E0033 — The Moment AI Stopped Being a Tool]] (2026-05-05) — application: agentic-AI cohort named as elastic automation extending from query to operate
- [[2026-E0035 - Copies of Copies/_metadata|E0035 — Copies of Copies]] (2026-05-07) — application: the constrained-task pile is exactly where Elastic Automators win, which is why the homogenization texture concentrates there
- [[2026-E0037 - I Caught an LLM at the Edge of Its World/_metadata|E0037 — I Caught an LLM at the Edge of Its World]] (2026-05-13) — **bridge body** with [[token-substrate-hypothesis|Cluster: Token-Substrate Hypothesis]]; uses *elastic automator* as the cold-probe term that the companion TSH paper formalized as the Coinage Probe across three frontier models.
- [[2026-E0038 - What the Phone Did to Work/_metadata|E0038 — What the Phone Did to Work]] (2026-05-15) — application: phone-era empirical dismissal of the "AI won't take your job, someone who understands AI will" trope; gestures at elastic automation as the system category the trope's vocabulary cannot name; grows-with-you accumulation is the load-bearing forward claim; close points at the elastic-automators concept page without naming the seed.
- [[2026-E0039 - The Word We Needed/_metadata|E0039 — The Word We Needed]] (2026-05-18) — companion-explainer arc opener to the EA position paper; the friend everyone knows scared of "AI" but not of the spreadsheet that holds his life; Ryle's Oxford-visitor illustration borrowed honestly to name the category mistake; "elastic automator" is the plain name the story hands over once the wrong word peels off; the ghost was never in the machine, it was in the sentence.
- (More Bodies will be added as the Arc continues.)
## Videos / diagrams / infographics
- E0029 short-form video: linked in the episode `_metadata.md` permalinks block.
- Future: paper-figure infographics; workflow-taxonomy diagrams.
## External references
- Anthropic Computer Use, OpenAI Operator, Cursor, Claude Code — the wild systems the framework names. Sources cited in the paper.
- Sutton, Barto — Reinforcement Learning (for contrast: rigid-action systems vs. elastic-language systems).
## Related topics
- [[ai-cognition|Cluster: AI Cognition]] — what cognition would actually require
- [[textual-kinematics|Cluster: Textual Kinematics]] — the physics-of-text view of the same generators
- Workflow intelligence (Cluster TBD as Bodies accumulate)
- AI agents / LLM tool use (Clusters TBD)
## FAQs
**Q. Isn't "Elastic Automators" just renaming AI agents?**
A. No. Renaming would preserve the consciousness-shaped framing. Elastic Automators is a *category change* — the framework is automation taxonomy, not agent taxonomy.
**Q. Does the framework apply to systems that include memory and tool use?**
A. Yes. The diagnostic-loop questions (*what loop, what tools, what memory, what criteria*) presume tool use and memory; they're how you map the system rather than reasons to call it a mind.
**Q. What about future systems that genuinely become minds?**
A. The framework explicitly carves out the boundary. When a system can pass the loop questions *and* satisfies the harder consciousness criteria (architecture-of-experience, not just behavior-of-response), the Elastic Automator framing no longer applies. The point is to avoid prematurely applying the mind framing to systems that don't earn it.
## Latest updates
- **2026-04-27** — Position paper v1.0.0 published on Zenodo.
- **2026-04-26** — Companion episode E0029 drafted.
- *(future)* — Diagnostic-loop-questions framework expansion paper.

