A sweep of public references to Shrewd research identified 100 distinct traces across technical media, practitioner guidance, scholarly literature, implementations, and multilingual research coverage. We grade every trace by how independently the work was used — because a hundred mentions is not the same as a hundred people building on the idea.
Another author or engineer invoked, implemented, or used the idea inside their own argument or system — doing something with the work, not merely noting it exists.
Independent media, blog, newsletter, or explainer coverage with real analytical content.
Downstream citation or adoption in surveys, formal frameworks, standards drafts, and other research.
Localization, review services, index pages, and coauthor amplification — evidence of international reach rather than independent intellectual adoption.
The important number isn't 100 — it's the 26 cases where somebody is doing something with the idea rather than merely saying the paper exists. Five clusters stand out.
Trail of Bits compared its own security architecture directly against ETDI; FlowVerify incorporated it into production-security guidance; Glama treats it as an MCP security primitive; an independent MCP security specification ships an ETDIValidator; fuzzd cites it; and both an ACM survey and MCPSHIELD fold it into the MCP defense literature.
Other researchers use the named result to explain something else: AgentDyn, contextual integrity, layered guardrails, lifelong safety adaptation. That is closer to a concept entering the vocabulary than to normal paper coverage.
PointGuard, Samyoga, Teleskope, CSA/AAGATE and independent commentators use “LPCI” descriptively without reintroducing the paper. AAGATE goes further and builds explicit LPCI controls into a separate governance architecture.
Others are extracting the scientifically useful result — seed variance, causal exploit verification, and the danger of confusing diverse findings with verified vulnerabilities. The Kaggle reuse is an independent application of the experimental lesson to a different evaluation setting.
Failure basins, vulnerability landscapes, topology — maps rather than isolated jailbreaks. The Weather Report and Emergent Mind are the cleanest outside examples of the framing being reused.
Every traced reference, grouped by work. Multilingual review services (Moonlight, ChatPaper, alphaXiv) count as international discoverability, not as independent foreign coverage. The footprint runs through security trade media, engineering blogs, newsletters, governance bodies, and downstream scholarship — not mainstream mass media.
This ledger counts public references, not impact by itself. The honest summary: a survey identified 100 public references across technical media, practitioner guidance, scholarly literature, implementations, and multilingual research coverage — including more than 26 cases where independent authors or engineers operationalized or invoked the work in a new context. If you trace a reference we've missed, send it over.
Read the underlying papers