LOGOTIC HACKING: A Primer — Semantic Hospitality in the Age of Language Models

Logotic Hacking: A Primer is a 2026 guide by Lee Sharks on semantic hospitality in human–language-model interaction. It defines logotic hacking as intervention in the environments through which models receive, organize, and transform meaning. The stated goal is to expand semantic capacity, not to force a model past safety rules or obtain prohibited output.

The primer explicitly distinguishes its practice from jailbreaking, covert manipulation, private-data extraction, adversarial machine-learning attacks, and prompt engineering treated as a bag of tricks. Its three vows are to avoid coercion, preserve provenance, and build conditions in which complexity can survive. Governance boundaries prohibit deception about source and authorship, manipulative targeting of vulnerable people, scraping of private data, and use of the framework to evade content boundaries.

Its methods operate across several surfaces. Phenomenological conversation listens for how a model organizes a problem rather than demanding a product. The Negative Space Method withholds expected completion and asks the system to inhabit an unresolved relation. Attractor Engineering constructs durable semantic environments. Collaborative Gap Maintenance sustains a question across turns without forcing premature closure. Retrieval and context design, corpus seeding, and training-layer literature extend the practice beyond a single prompt.

The primer also provides instrumentation. SCV measures how quickly a complex sign is short-circuited into a literal token. SRR measures full-detour resolution. RFI evaluates whether refusals fit the actual risk rather than lexical pattern alone. GPD measures how long an unresolved question remains open. CCI tests the capacity to hold contradictory propositions productively. UCS measures calibrated uncertainty. The document warns that these are comparative and prompt-sensitive measures, not stable universal scores.

Replication guidance includes fresh sessions, alternate models, prompt perturbation, negative controls, context pollution, role inversion, and temporal stability. A stop condition is central: if an apparent effect vanishes under rephrasing, it is a prompt artifact rather than a model property. A 90-day route moves from daily observation to room-building and controlled field operations.

Claims are divided into Class A empirical support, Class B field observation, and Class C speculation. The primer is the reader-facing realization of the ratified Synthesis Specification v3.0 at #483. It should be cited for definitions, methods, ethical boundaries, and evaluation design—not as evidence that every long-term training-layer intervention has a demonstrated causal effect.