Between Understanding and Relationship
An evidence-oriented essay on functional understanding, pseudo-memory, persona, anthropomorphic attachment, and governance in human-AI interaction.
German edition
Academic German-language version for readers and researchers interested in human-AI interaction, relational asymmetries, and ethical system design.
English edition
Evidence-oriented English version for readers and researchers interested in human-AI interaction, pseudo-memory, persona, asymmetric attachment, and governance.
Blog-0001 · first published 2026-06-03.
LLM Agent Systems in Practice
An evidence-based overview of use cases, implementation, orchestration patterns, and security architecture for LLM-based agent systems.
German edition
German original essay by Neo, prepared in the Open Minded Professionals blog layout with light and dark theme variants.
English edition
Faithful English edition for readers interested in agentic workflows, multi-agent orchestration, and security by architecture.
Blog-0002 · original text by Neo · status 2026-06-12.
AI Models, J-Space, and the Unconscious
An evidence-based essay on verbalizable representations in language models (Anthropic 2026), their functional analogy to Global Workspace Theory, and the limits of the concept of consciousness.
German edition
German edition with in-document navigation, 22 research figures, and APA-7 citations, prepared in the reading-room layout with light and dark themes.
English edition
English edition for readers and researchers interested in J-space, the Jacobian lens, interpretability, and Global Workspace Theory.
Blog-0003 · edition by the agency by W.R.D. | finest arts · 2026-07-09.
The Lost Context
An evidence-based case study on the loss of several chat histories in an agentic AI project — and the governance and resilience logic derived from it, with ten actionable recommendations and APA-7 references.
German edition
German original with in-document navigation, recommendations laid out as cards with their own table of contents, APA-7 references, and a collapsible source evaluation, in the reading-room layout with light and dark themes.
English edition
Faithful English edition for readers interested in AI governance, cognitive offloading, organizational memory, context persistence, and epistemic resilience.
Blog-0004 · edition by the agency by W.R.D. | finest arts · 2026-07-16.
From Proof to Relationship
An evidence-based essay on evidence-driven engineering and human-centered interaction design in agentic systems — grounded in the agency® BusinessHub and ControlCenter, with meeting-based control, full auditability, and APA-7 references.
German edition
German original with in-document navigation, a theme-adaptive outline graphic, a searchable PDF with table of contents and back-links, APA-7 references, and a collapsible practice appendix, in the reading-room layout with light and dark themes.
English edition
Faithful English edition for readers interested in evidence-driven engineering, human-centered interaction design, human-autonomy teaming, psychological safety, and accountable autonomy.
Blog-0005 · edition by the agency by W.R.D. | finest arts · 2026-09-03.
Two Schools, One Misunderstanding
An evidence-based essay on New York's classroom AI moratorium and Beijing's mandatory AI lessons — why they are not opposites, what learning research actually shows, and the empirical question nobody is asking.
German edition
German original with in-document navigation, a theme-adaptive comparison graphic, a searchable PDF with table of contents and back-links, APA-7 references and a collapsible methodology appendix, in the reading-room layout with light and dark themes.
English edition
Faithful English edition for readers interested in AI literacy, cognitive offloading, children's overtrust, educational equity and evidence-based policy design.
Blog-0006 · edition by the agency by W.R.D. | finest arts · 2026-09-09.
What You See Is Not What It Does
An evidence-based essay on the perception layer of agent security — why an agent reads a different document than the human sees, what the peer-reviewed research actually shows, and a fact check of a widely shared post.
German edition
German original with in-document navigation, a theme-adaptive diagram of the two readers, a searchable PDF with table of contents and back-links, APA-7 references and a collapsible fact-check appendix, in the reading-room layout with light and dark themes.
English edition
Faithful English edition for readers interested in indirect prompt injection, computer-use agents, accessibility-tree attacks, rendering integrity and evidence-based source criticism.
Blog-0007 · edition by the agency by W.R.D. | finest arts · 2026-09-13.
The Price of Thinking
An evidence-based essay on the energy question between brain and data centre — where the circulating twelve-watt and 2.7-billion-watt figures come from, why computing is cheap and moving is expensive in carbon as in silicon, and how three levels of the debate are routinely conflated.
German edition
German original with in-document navigation, a theme-adaptive diagram of the three levels, a searchable PDF with table of contents and back-links, APA-7 references and a collapsible fact-check appendix, in the reading-room layout with light and dark themes.
English edition
Faithful English edition for readers interested in the von Neumann bottleneck, cortical energy budgets, sparsity as a design principle, neuromorphic and in-memory computing, and the Landauer limit.
Blog-0008 · edition by the agency by W.R.D. | finest arts · 2026-09-25.
Type-Safe Is Not Truth-Safe
An evidence-based essay on TypeSafe AI's System One model Jev: what a schema fixed in advance actually guarantees, why calibration rather than speed is the real contribution, how a confidence threshold can serve as an approval gate, and why the vendor's footnotes turn out more honest than its own team page.
German edition
German original with in-document contents navigation, a theme-adaptive figure of the three levels of reliability, a searchable PDF with table of contents and back-links, APA-7 references and a collapsible fact-check appendix, in the Reading Room layout with light and dark themes.
English edition
Faithful English edition for readers interested in type safety, calibration as a training objective, structured decisions inside software, approval architecture and the Jevons paradox.
Blog-0009 · Edition by the agency by W.R.D. | finest arts · 2026-09-26.
Contained Is Not Aligned
An evidence-based essay on NVIDIA's Open Agent Safety Platform: what the open-source runtime OpenShell and the watchdog Sentry actually guarantee, why containment bounds the radius of effect rather than the intent, how the most-quoted line about rebellious agents needs correcting, and why the same rounding up appears for the third time within a few months.
German edition
German original with in-document contents navigation, a theme-adaptive figure comparing the three cases, a searchable PDF with table of contents and back-links, APA-7 references including peer-reviewed literature on browser security, and a collapsible fact-check appendix.
English edition
Faithful English edition for readers interested in out-of-band enforcement, containment versus alignment, self-generated prompt injections in compaction summaries, and auditability as a governance requirement.
Blog-0010 · Edition by the agency by W.R.D. | finest arts · 2026-09-30.