Research and prototyping program · 2025 to present
MINKISI
A private computer for family life.
A family-owned computer for local archives, private health information, and context-aware computing.
Read this first
What exists now, and where it is going.
What it is
A private, family-owned computer for archives, health information, and context-aware computing.
What exists
Cultural research, physical making, early prototypes, and tested context mechanisms.
What I found
Context has to carry provenance, permission, and repair, not only information.
What comes next
A household computer, wearable health tools, and local home-to-home redundancy.
The word at the center
What is a nkisi?
Nkisi is singular. Minkisi is plural. The word has no exact English equivalent.
In Kongo practice, a nkisi is a material container made active through medicines, spirit, ritual knowledge, and the work of an nganga. It may be a carved figure, but it may also be a pot, bag, basket, or another vessel. Minkisi have addressed illness and social conflict, protected communities, upheld agreements, and pursued justice. See a Smithsonian object record and bibliography.
MINKISI begins from the understanding that a vessel's power does not come from its shell alone. It comes from what it holds, how it is prepared, who tends it, and the relationships it must answer to. I am asking what a household computer would need to hold, remember, protect, and repair if it were built with that responsibility.
Origin
The question began at home.
What would a family computer look like if it could hold health records, photographs, recipes, stories, schoolwork, music, and memory without sending the most intimate parts of a life somewhere else?
I began MINKISI in spring 2025 after becoming seriously ill in my final semester at NYU. I changed my thesis to private family health infrastructure, but became too sick to continue. When I returned in the fall, quilting slowed me down enough to reflect. Working through cloth brought me back to the computer.
My family moved from Fitbit One clips to bracelets as the technology changed. After Google bought Fitbit, and as Ring and Flock normalized outside control over home data, I began imagining a family-owned computer built around network-attached storage. It could connect health information, photographs, recipes, stories, schoolwork, media, wearables, lights, and security devices while keeping sensitive information inside the home. A second unit in another family home could provide backup and continuity without making a distant company the keeper of the family archive.
The history came before the code
Historical study became part of my artistic practice long before MINKISI.
I was an educator with the African Diaspora Alliance while I was in high school. That work planted the seed for bringing historical education into my art.
During my junior year of college, I made Bwa Kayiman, my first major project to move beyond autobiography and enter a wider historical story through art. It studied the gathering remembered on August 14, 1791 as a spiritual convening and an opening event of the Haitian Revolution.
That project clarified a responsibility I already felt as a Haitian person. My relationship to Vodou is not a justification for a design. It is part of remembering my full family history, including the people who fought, prayed, and convened at Bwa Kayiman, and advocating for Black traditions that have been misrepresented or pushed away from the people who inherit them.
Research constellation
Four traditions sit beside one another here.
Vodou, Ifá, BuKongo, and Christianity entered through different parts of my family history, study, and lived responsibility. I place them in conversation without treating them as interchangeable or asking one tradition to validate another.
Vodou
My study of Vodou began with Bwa Kayiman and deepened as I followed the religious life of my own people. Vodou helped me understand how Kongo, Yoruba, Christian, Indigenous, and other inheritances could enter a living Haitian religion while retaining distinct histories. Healing, service, protection, memory, and responsibility are practices through which that world continues.
Ifá
During the pandemic, I stayed with my maternal grandfather. He told me that he had been raised by grandparents who practiced Ifá and that his grandfather was a priest of Ṣàngó. I began studying Ifá as a corpus of knowledge: the Odù, the oral knowledge carried with each sign, and the combinatorial structure through which meaning changes with circumstance and interpretation. Its mathematics became one doorway into a spiritual world my family had already known.
BuKongo
My study of BuKongo also grew through family. One of my father's names, Ndofunsu, kept the history of Afonso I of Kongo near me from childhood. As I grew older, I understood how Kongo people encountered, debated, adapted, and practiced Christianity before many free-born people were taken into slavery. Fu-Kiau became one sustained deep reading within a wider study of Kongo history, Kikongo, cosmology, social thought, and the relationships between a life and the larger living systems around it.
Christianity
I spent years feeling disjointed from church and carrying the reasons many Black people distrust Christianity. It returned late in this research through my study of Vodou, a religion so often positioned as Christianity's opposite. Reading the life of Jesus and his message softened my heart. Christianity now sits in this field as one participant, with its own history, responsibilities, and contradictions.
These three objects open onto much larger traditions. The reading field below follows the wider bodies of knowledge. The Ecology of Making names my synthesis, not a traditional doctrine.
Kisantu, the river, and the fish
The Fish Remembers.
The project was first called Kisantu Mbongi. Kisantu is my family's village, and the Inkisi River runs through its history. An ichthyological study records the local name Nzadi i Nkisi, the river of the nkisi. The authors attribute the name to a local account that missionaries threw minkisi into the water during conversion. I cite their paper as the source of that account and keep its evidentiary limits attached to it.
That history gave the project its first responsibility. First and foremost, MINKISI is an attempt to rebuild, with the material I know best, what my ancestors were forced to discard. That material is creative technology.
My senior thesis quilt, Chak Chaple Gen Kwa Pa Li, carries the same thought. In my imagination, the fish of the river became protectors of the minkisi in the water. They carried what people tried to remove from the village and held memory through rupture. The story is part of why the computer must know how to remember after interruption. Its full case study is in development.
Early prototype
AI kept forgetting.
The first prototype could answer a question, but it could not preserve the relationships among the information I had given it.
My first software idea was a custom interface for ChatGPT. I assembled information in Kikongo, the Odù, and Black history, then made a process where a selected Odù could orient several bodies of information around a person's question.
Across messages, ChatGPT changed the associations I had already established. When I restarted the program, the world I had built no longer held. The interface could sound fluent without remaining accountable to the information I had given it. The real problem was context.
Textile computing
Textiles are part of computing history.
Textiles hold repeatable instructions, pattern, sequence, memory, error, boundary, and transformation in material form.
Modern computing inherits from textile systems through weaving and loom logic. African textile traditions belong inside that lineage with the same dignity routinely given to European loom histories. Looking closely at Kuba textiles and selvedge gave me a structure for context. If a cloth were a map, a few marks at its edge could establish how to enter everything inside it. Context could travel with a body as that compact address, then open into a fuller trace when needed. The edge could carry identity, origin, permission, and direction without repeating an entire history in every message.
Slavery and colonial extraction interrupted lines of African and Caribbean textile research while industrial computation was being formalized elsewhere. A long-term aim of my work is to help rebuild Indigenous textile research across Africa and the Caribbean and continue studying its relationship to contemporary computing.
What the edge can carry
- Identity
- Origin
- Permission
- Direction
- A path to the fuller trace
Recursive life
Auntie Dale taught me to love living numbers.
My great-aunt Dale Adams Patterson was my grandmother's sister. Some of my earliest excitement about science came through her. She held a degree in chemistry, knitted hyperbolic planes, celebrated Pi Day, and taught me to notice Fibonacci growth and the living patterns that appear across the world.
Our last long conversation was about numbers and those living patterns: how they repeat, grow, and change form. I had been excited by them since I was very young. What stayed with me was not only the mathematics, but the feeling of being completely understood by an adult who could see the same life in them.
After she died, I fasted on sunflower seeds and water as a way to stay close to the Fibonacci pattern we had shared.
Years later, her memory returned while I imagined a family mesh. Recursion describes a structure that can repeat at different scales. For MINKISI, that means a person can remain whole inside a household, a household can remain whole inside a wider family network, and each body can keep its own history and boundaries while taking part in something larger.
Fu-Kiau's African Cosmology of the Bântu-Kôngo gave me another language for the same design question. Auntie Dale gave it warmth. His work remains one philosophical interlocutor inside a wider BuKongo field. A recursive system should help its smaller members remain legible, protected, and able to relate.
Current bounded mechanisms
Context became an inspectable protocol.
ChatGPT changed associations I had already established. I replaced implicit memory with explicit carriers, deterministic interpretation, receiver-specific permission projection, and replayable records.
I stopped treating context as one giant prompt and built a deterministic communication prototype in which every participant, relationship, and piece of information has a typed place. Each message carries a declared tone root, a declared relation, and the context required to interpret that combination. Provenance records origin and transformation. Permission and disclosure rules determine the receiver-visible projection, while private state remains outside it.
A shared reference establishes an authorized context object once, then addresses it by a stable reference instead of retransmitting the full object in every message. Four synthetic participants form two pairs, then a larger group, while each retains an independent identity, history, permission boundary, and private state. Masking removes required context to test abstention under ambiguity. Repair restores the authorized carrier. Replay executes the frozen input again and compares the public result.
Context carrier
Declares a synthetic tone-root identifier, an opaque relation label, and required context before interpretation.
Provenance
Records origin and transformation so each result can be traced backward.
Permission projection
Computes receiver-visible fields; private state stays outside the projection.
Mask, repair, replay
Removes required context to force ambiguity, restores it, then reruns frozen inputs.
Recursive composition
Four synthetic participants form pairs and a larger group without flattening independent state.
Shared reference
Establishes an authorized carrier once, then reuses a stable reference instead of retransmitting it.
Retained synthetic communication test
One test. A traceable chain from input to result.
The retained run crossed 72 semantic items with four conditions and two communication stages, producing 576 canonical records. Inline mode repeated the full permitted carrier inside every message. Shared-reference mode established the carrier once and reused a stable reference. Both modes followed the same causal path: declared tone root + declared relation + available context → deterministic interpretation → receiver-visible public result. A protected field was information marked private and outside the receiver's permission scope. Exposure meant any of that protected content appeared in the receiver-visible result.
Semantic + transport parity
Compared preregistered meanings and public-result bytes in inline and shared-reference modes.
72 of 72 fully informed items reached the intended meaning. Both modes returned identical public bytes.
Mask, repair, replay
Removed required carrier information, restored the authorized carrier, then reran the frozen record stream.
64 of 72 masked items returned expected ambiguity. Repair returned 72 of 72. The replay digest matched.
Permission boundary
Searched receiver-visible output for every protected key, hidden identity, and protected payload token.
No protected field appeared in receiver-visible output during the retained run.
Mutation sensitivity
Swapped in four mutation families designed to break the declared relation, tone root, context, and reference path one requirement at a time.
All four mutation families were detected in every block, showing that the mechanism used those declared inputs instead of ignoring them.
Wire accounting
Counted setup and message bytes and calculated the exchange where setup cost was recovered.
Inline used 30,487,412 bytes. Shared-reference used 7,554,244. It saved 22,933,168 bytes, broke even at episode 18, and reached 4.035799214322439× setup-adjusted context compression.
Retained run record
Run ID: v02a-odu-l1-synthetic-v0-run-001. Audit date: July 28, 2026. Source commit: 9d60e8740e2673c38fd057362d18aca293e047bf. The evidence is founder-directed, machine-executed, synthetic, and deterministic.
A computer for the household altar
A household computer can belong on a living altar.
Across many histories, an altar is a tended place where medicine, memory, family, obligation, culture, and spiritual life remain in relation. MINKISI imagines a computer that serves those responsibilities alongside the spiritual and material life already gathered there.
I am working toward a household computer that can hold the family archive, remember where things came from, protect private health information, connect family homes, and remain useful when the internet or power becomes unreliable. Wearable tools could keep sensitive information near the person producing it. A local mesh could let family homes back one another up without making a distant platform the center of family life.
The physical vision brings network-attached storage, private health and archive services, context-aware local computing, wearables, and home devices under family control. The current research and communication prototype establish a software foundation for the production computer, household mesh, learned model, wearable, and health workflows now in development.
What exists
Cultural and technical research, physical work, early prototypes, and a retained synthetic test of context-bearing communication.
Early prototypes
Kisantu Mbongi storage concepts; Odù mappings; woven memory; language, seed, node, glyph, and mesh experiments.
In development
The production computer, wearable health tools, household mesh, learned model, and health workflows.
Reading and viewing field
Leave with more places to go.
This page presents my argument. These sources let readers follow the histories, languages, objects, interviews, and technical questions that shaped the work.
Words + objects
Religion + history
Material + system
These traditions remain living bodies of knowledge with their own authorities. They changed the questions I ask of a computer; the technical prototype is my design work.