ETHRAEON / Papers / Essays / The Reflective Architecture

The Reflective Architecture

Architecture before features.

S. Jason Prohaska (ingombrante) · Bologna, Italy · 7 September 2026 · Published on Substack · CC BY 4.0 · Attribution required to S. Jason Prohaska (ingombrante).

Last week I had the surprising opportunity to meet Paolo Zuffa in Bologna at Room with a View in Piazza Santo Stefano via an introduction from a close friend, Luigi Falchieri.

Paolo Zuffa

Paolo is an Italian-qualified lawyer with nearly thirty years of international business-law practice EU antitrust, M&A, and cross-border regulatory work in London, Brussels, Moscow, and Tbilisi who since 2019 has redirected his work toward the regulatory analysis and design of conversational AI.

In August 2026, publishing under the name "Humanising AI," he released The Reflective Architecture, an open, non-commercial design framework that proposes dependency-resistance rather than engagement or task-completion as the primary design constraint for conversational systems operating in emotionally intimate contexts.

Professional Timeline

2019 - present. Independent Researcher, Regulatory Analysis of Conversational AI Design (self-employed, Italy). Developed The Reflective Architecture and its companion specifications; research areas include memory architecture, user sovereignty, arc-neutral interaction design, intentional forgetting, and distress-response protocols.
2015 - 2019. Independent Business Consultant, Tbilisi, Georgia. Advised an Italian general contractor on World Bank and Asian Development Bank procurement tenders for road-construction projects, and on a public tender for a waste-management facility.
1998 - 2014. Independent Business Consultant, Moscow, Russian Federation. Advised a major Russian metals group on the acquisition of an Italian steel-smelter group; advised an Italian metals company on building a Russian raw-material supply chain; advised an Italian rail-control systems producer on the sale of its product line to Russian Railways (RZD).
1996 - 1998. Associate, Gide Loyrette Nouel, Moscow. M&A and Russian antitrust law during the post-Soviet transition.
1992 - 1993. Associate, Slaughter and May, Brussels. EU antitrust law - competition-law compliance, merger control, regulatory strategy.
1990 - 1992. Associate, Slaughter and May, London. M&A and UK company law - asset and share purchase agreements, joint ventures, distribution agreements.
1985 - 1988. Ricercatore (Research Fellow), University of Rome II (Tor Vergata), Rome. Lectured in International Law and EU Antitrust Law.

Source: CV_Paolo_Zuffa.md, humanising-ai/reflective-architecture repository.

Luigi and I

We met five years ago when we both moved to Bologna. He was returning home after working in London for over 11 years. He returned to Emilia for many reasons during the Covid pandemic, including a desire to have his son, then in 5th grade, experience and grow up in a city and community far from British hooliganism, and to be closer to grandparents and family. His son and my son also enrolled in fencing together at Virtus Academy in Bologna. Luigi's son continues to compete, and he met Paolo at a competition, as Paolo's son is three years ahead, both in school and in his fencing endeavors.

A bridge to dinner and a (quiet) Room with A View

A few weeks ago over dinner, Luigi and Paolo were discussing their occupations, which in Italy is rare to chat about work life over a friendly meal. Paolo is a retired attorney, having worked internationally his entire life, and he also returned to his childhood home in Bologna around the same time in 2021, for very similar reasons.

Over dinner, Paolo discussed a theoretical framework he had derived from decades of jurisprudence and his experiences with LLMs and AI. He shared a Git repository, and as Luigi has been an advisor to me since March 2025 in my development of Ethraeon, he noted many similarities between Paolo's theories and my evolution of Ethraeon.

As a result, last week Luigi sent me a text, shared the details, and asked for a reading of what Paolo had derived and put forward.

"Hey Jason, please check this link github.com/humanising-ai/reflective-architecture and let me know what you think / the guy who developed this, Paolo Zuffa, is Costantino's dad. I met him at a scherma competition. - Gigi."

And I accepted.

I engaged my systems and produced two valuable artifacts, co-authored by me and my systems. I did not request my systems ingest the Git, for obvious reasons, however I did state the intent of evaluating the full repository, and why.

In about :20m the first artifact was created, which was produced for the intent of passing back to Luigi and, per his request, with his discretion to share with Paolo if he chose to do so.

After I shared the link to the analysis, I decided to push further and use my systems, leading with LUMENA, to bridge Paolo's theory and build a sovereign kernel that demonstrates his intent with the capabilities of my systems. This effort produced the second artifact.

I shared both on text with Luigi, approximately :40 minutes after he had first texted me with his request.

"Ciao Gigi. Due Spuntini to read and also share with Paolo if you wish. It's remarkable the alignments and conclusions he has formulated in contrast with the maturity of Ethraeon & my systems. Let's get together if you want, I'd also love to visit with Paolo if he's keen.

1. analysis.ethraeon.ai

And I created a prototype. Give it a try:

2. analysis.ethraeon.ai/rdk/

- Jason"

Luigi responded simply:

"Can I forward it to Paolo? And you guys, if interested, can take it from there? - Gigi"

I of course approved, and a few days later I received a follow-up text from Luigi:

"Ho ricevuto un messaggio da Paolo... che mi chiede se ci incontriamo per un aperitivo domani alle 19:00, in Piazza Santo Stefano... Camera con Vista. Ci saranno anche Valentina e Tikka ( la moglie di Paolo ) che ne dici? - Gigi"

Translation:

I received a message from Paolo... he requests we meet for a drink tomorrow at 7pm at Room with a View in Piazza Santo Stefano. I will bring Valentina, and also in attendance will be Paolo's wife, Tikka. Whaddya say? - Gigi

My simple reply:

"Ci sto."

Translation:

I'm in.

Dinner and A Core Agreement

We met and enjoyed four hours of conversing about life and ideas and of course The Reflective Architecture and Ethraeon Systems, over three bottles of Sauvignon and appetizers.

Paolo smoked American Spirit cigarettes.

Gigi contributed wonderful stories of his technology inventions.

Valentina and Tikka chimed in between bites.

The core agreement: Architecture before features.

A wonderful evening and serendipity prevails.

Paolo and I will certainly collaborate again soon.

I find more and more there are no coincidences, and yes, as they say... "The world is an oyster."

What a ride.

And do try (a quiet) Room with A View. The latency is low but the replies are temporary and uniquely not therapy.

#gratitude.

Source Notes

  1. humanising-ai/reflective-architecture, by Paolo Zuffa. github.com/humanising-ai/reflective-architecture
  2. The Reflective Architecture · An Analysis by S. Jason Prohaska & Ethraeon Systems. analysis.ethraeon.ai
  3. The Reflective Dialogue Kernel, by S. Jason Prohaska & Ethraeon Systems. LUMENA x bridge using Paolo Zuffa's theory. analysis.ethraeon.ai/rdk/
  4. Prohaska, S. Jason (ingombrante). Time, Destiny, and the Destroyer of Worlds. Bologna, Italy, 23 May 2026. CC BY 4.0.
  5. Prohaska, S. Jason (ingombrante). Temporal Inversity and Ethraeon Systems: A Tenet Alignment with Ethraeon. Bologna, Italy, 30 August 2026. CC BY 4.0.

How to cite. Prohaska, S. Jason (ingombrante). The Reflective Architecture: Paolo Zuffa. The Reflective Dialogue Kernel. And A (quiet) Room with a View. Article (published). Bologna, Italy, 7 September 2026. CC BY 4.0.