Cognitive architectures
How AI remembers, reasons, and learns over time — seven-layer memory architecture (ZenBrain), sleep consolidation, neuromodulator-driven consolidation, FSRS-driven review.
Principal Investigator
Alexander Bering
Independent Researcher · ZenSation Research · Kiel, Germany
Research on cognitive architectures, AI safety in public spaces, and applied AI for knowledge work. Self-funded without external third-party funding. Open access on arXiv and with DOI on Zenodo; code under Apache 2.0. Alongside the research, he partners with organisations on methodologically grounded AI adoption — as a research partner, not a vendor.
The person behind ZenSation
The road to the architecture
Why a practitioner from foreign-exchange trading now pursues foundational research on machine memory. In his own words.
2009 — The first system failure that wasn't one.
I started automating foreign-exchange trading in 2009. Anyone who builds trading systems learns an uncomfortable lesson faster than any theory could teach it: the market punishes every assumption you have not made explicit. A system that worked yesterday and loses money today has not changed — it just never understood why it worked. That experience has permanently shaped my relationship with software: I do not trust a system that cannot remember its own decisions.
From trading to systems thinking.
Out of trading automation grew the real question: how do you build systems that learn from their own history instead of forgetting it? Years of practice inside large corporations — at the seam between grown processes, real data landscapes and what software can actually deliver — sharpened that question. Corporations forget just like trading systems do: in inboxes, in departing heads, in project drives nobody opens any more.
Cognitive architectures.
Today's answer to that question is collective intelligence on cognitive architecture — humans and machines together, on a structure that retains knowledge. Not a bigger model, but a better structure around the model: a memory that carries decisions, reasons and context across years. A memory that belongs to someone else is none. It must belong to the one whose experience it holds.
The paper.
This work is published: ZenBrain (arXiv:2604.23878) describes the architecture in full — 15 memory mechanisms, 9 of them foundational, 6 as PMA. A package of 20 modules is open source. I publish what I build because verifiability is the currency that counts in this discipline.
Kiel.
ZenSation works out of Kiel. That is neither an accident nor a disadvantage: foundational work needs concentration more than proximity to conference stages. Whoever works with me works with the person who built the systems — there is no second row to delegate to.
Identifiers
Academic profiles and identifiers
Research focus
Three research tracks
Safety in public spaces
Predictive early warning in public spaces under a civil-liberties architecture (CrowdGuard) — explicitly without biometric identification, helper protection as a structural element of the objective function.
Applied AI for knowledge work
Research on human–AI collaboration in organisations — meetings, project management, knowledge work, and process mapping, including a methodology for structured AI adoption. At its centre is ZenAI as a productive research application — a vendor-neutral AI operating system on the open core ZenBrain, where we test whether the memory research holds up in daily use.
Background
Professional and research-biographical background
- Since 2009
Fully automated, algorithmic trading system (MQL4 in MetaTrader 4), live on the personal computers of thousands of customers — with dedicated research and systematic testing. Methodological foundation: statistical modelling, deterministic strategies, real-time processing.
- 2012
Work on digitalisation projects at scientific institutions — among others at the IPN – Leibniz Institute for Science and Mathematics Education at Kiel University, and at IQSH (Institute for Quality Development in Schools Schleswig-Holstein), on the digitalisation of learning and educational materials.
- 2010s–2020s
Digitalisation and transformation projects at the corporate and stock-corporation level: process atlases, AI strategies, and AI-supported decision systems across industry contexts.
- Security & operations
Hands-on operational experience in security — on active duty as an emergency responder and in the commercial security sector, including as a close-protection officer. Alongside this, work for German public-safety authorities and organisations (BOS) and the Federal Ministry of the Interior (BMI), partly voluntary and partly project-based — including in incident-sector command and in committee work. This operational reality shapes the applied orientation of the public-safety track.
- 2024
Reorientation toward fundamental and applied research. First form of the cognitive architecture with reliable long-term memory; in parallel, the question of fundamental-rights-preserving safety mechanisms for public spaces.
- 2025
Conception and first implementation. Establishment of the three research tracks (cognitive architectures, safety in public spaces, applied AI for knowledge work). Build-out of the research documentation, pre-registration of central documents via OpenTimestamps.
- January – March 2026
Production implementation in the internal repository: 7-layer memory architecture, test suite, and the first application (ZenAI), architectural consolidation across multiple phases. Late March: open-source release of the ZenBrain packages on npm under Apache 2.0.
- May 2026
Publication of the ZenBrain paper as an open-access preprint on arXiv (2604.23878) and with a permanent DOI on Zenodo (10.5281/zenodo.19353663) — publicly accessible, with replication material.
- In preparation
Follow-up publications for peer-reviewed venues. Exploratory exchanges with universities and research consortia across the European research area on consortium proposals along BMBF SIFO, Horizon Europe Cluster 3, and national funding lines.
Research collaborations
Scientific exchange
Research in an independent lab does not occur in a vacuum. Outreach to potential co-authors is consistently kept in conditional terms — names appear on this page only after written consent. This discipline protects potential partners and distinguishes exploration from established collaboration.
Outreach in progress
Outreach to research institutions across the German and European space is ongoing in waves 2026–2027. Cooperation partners are named publicly only after receipt of written consent.
Three rules that decide everything here
Evidence before assertion.
What we say about our systems is published, disclosed, or verifiable live in the playground. What we cannot substantiate, we do not claim.
Transferability as proof.
Good work can be recognised by the fact that it can be continued without its author. Every ZenSation project therefore ends with a handover, not with a dependency: you remain capable of acting, even without us.
Research funds itself.
ZenSation takes on mandates so that the foundational work stays independent — not the other way round. Every euro from projects flows back into open, verifiable research.
If you want to talk to me
For companies considering a collaboration, the Working-with-us page is the right entry point. For scientific enquiries, reproductions and data access: research@zensation.ai. I answer both personally.
Contact
Inquiries are typically answered within ten working days.
More from this research
Related pages
Core and application fields, on a shared ethics foundation.
- Research Three tracks: cognitive architectures, safety in public spaces, applied AI.
- Technology ZenBrain in depth — 7 memory layers, algorithms, RAG pipeline, performance.
- Adopt AI A neutral roadmap for AI adoption — prioritisation, stage-gates, KPIs and EU AI Act / GDPR governance, with an interactive simulation.
- Methodology Pre-registration, reproducibility, external validation, data minimisation — operational standards in detail.
- Publications arXiv, Zenodo (DOI), software releases, open-access principles.
- Benchmarks ZenBrain by the numbers: every figure with its method, source and verification link.
- Resources Code, replication material, BibTeX citation, licences, identifiers.
- Research ethics GDPR Art. 89, AI Act Art. 5, Brokdorf line. Eight mandatory corrections.
- Public sector & funding Research offerings for BMBF, BBK, universities and research consortia.