Prolegomenon
Common threadWe start from a real person with a real problem, and make complex domains readable and verifiable without losing what matters.
Law, accounting, archives, sacred spaces, speech, mountain trails: every domain we work in has an order of its own. That order is real, but hard to read. The people who live inside it rarely see it as a whole.
Systems theory, since Luhmann, holds that a system exists to reduce complexity. We take the opposite view: make it readable without making it poorer. Our work is to build the tools that let people move through a complex domain, question it at any point and check every step.
Three operations run through everything we do:
- Mapping. We reconstruct hidden structure from scattered sources. A map as large as the territory is useless, as Borges noted. A good map leaves things out on purpose.
- Translation. We carry meaning from one register to another: from law to citizen, from an old system to an archive, from a place to a story, from speech to text. What made the original reliable must survive the journey.
- Proof. Nothing is true because someone said it, and nothing is finished because someone declared it finished. As Popper argued, a claim holds only while it survives the attempt to disprove it.
The fourteen fields below are disciplines of computer science. Law, finance, heritage and rhetoric appear in them only as areas of application. Each field is stated in a few theses and ends with an aporia: an open question we have not solved. That is where the conversation starts.
Origin
Most fields on this page began with a problem that someone in our team met first-hand: a tax return in a new country, a back that hurt, a trail that no map showed, a speech that had to get better, a parish that wanted to be seen. The others began with a client's problem that we made our own. None was chosen from a market report. We built what we needed, with the rigour we would expect as clients, and then opened it to others. That is why our answers are our own.
Behind every system there is a person who is tired, worried or in a hurry. We design for that person first.
IEnterprise Information Systems and Software Architecture
Common threadThe life of a company, recorded so that it can be checked at any time.
Vitruvius asked every building to be solid, useful and beautiful. An enterprise system must also be accountable.
I.1Compliance belongs in the architecture, not in the procedure. Legal retention, tamper-proof records and audit trails are built into the core. Rules that depend only on discipline are eventually broken for convenience.
I.2State is calculated, not stored. Every event is recorded with its date, and the current status is computed from that history. This is Pacioli's double-entry bookkeeping (1494), applied to everything a company owns. Mistakes are corrected by adding a new entry, never by deleting the old one.
I.3Money is not just a number. Currencies, rounding rules and tax regimes have their own logic. Treating money as an ordinary decimal is a design error, not a shortcut.
I.4Reuse is earned, not predicted. A pattern is first proven in real operation. Only then is it carried into other products.
One system can serve the owner, the employee, the citizen and the merchant. How many roles can it carry before permissions start to leak from one to another?
Grounded inan operational ERP in industrial manufacturing; a multi-tenant platform for Swiss SMEs (in trial).
IIProcess Automation and Robotic Process Automation
Common threadRepetitive work handed to machines; judgement kept by people.
Čapek coined the word robot in 1920, from the Czech robota: drudgery. Good automation takes the drudgery and leaves the judgement to people.
II.1No interface? The screen becomes the interface. When an old application has no programming interface, its screens are first described precisely: menus, dialogs, shortcuts, reports. Only then is the first step automated.
II.2Meaning before pixels. When a screen exposes its structure, the robot works with that structure: fields, roles, states. Clicking on pixels is the last resort.
II.3The scenario is the asset; the robot can be replaced. A business scenario is written once. It can then be run as a guided overlay for a person, through remote assistance, by voice, or automatically in the background.
II.4Automation without a stop button is abdication. Every robot needs consent, a stop that always works and a record of what it did. What it must never touch is blocked by design, not left to judgement.
Suppose the pricing logic of a closed old system is rebuilt from the reports it prints. Can the copy be good enough to replace the original, and who answers for the difference?
Grounded ina guided-scenario engine running on an industrial ERP; automation of a desktop application that has no programming interface (in trial).
IIIData Engineering, Migration and Archival Science
Common threadData that outlives its systems and always shows its source.
Derrida wrote that every archive is torn between the wish to keep and the urge to destroy. Engineering has to choose.
III.1Systems die; their data must not. Archiving means making information outlive the software that created it.
III.2No source, no answer. Every record we return names where it came from. "Unknown" is an acceptable answer; an invented one is not.
III.3Structure is discovered, not assumed. Undocumented databases reveal their relationships through analysis of the data itself. Old documentation rarely tells the truth.
III.4A good digital copy travels without changing. A scanned page becomes searchable and portable, and keeps the layout of the original. Latour called such objects "immutable mobiles".
Can the meaning of an undocumented database be reconstructed reliably through checks alone? Or will an expert always be needed?
Grounded inan archiving service for legacy enterprise systems (in trial); a step-by-step migration method for an insurer's core system.
IVInformation Retrieval, Knowledge Representation and Legal Informatics
Common threadThe law made searchable, and every answer tied to its article.
Leibniz dreamed that disputes would one day be settled with the words "let us calculate". The whole dream is out of reach, but part of it can be built.
IV.1A law is a network, not a text. It is made of references, amendments and dates of validity. It can be searched properly only when it is treated as linked data.
IV.2An answer without a citation is a rumour. An answer drawn from the law is worth exactly as much as its ability to point to the article behind it.
IV.3Civic knowledge comes in three steps: inform, learn, act. The path leads from the official term, to its explanation, to the reader's own case. In a country with twenty-six cantonal tax systems, Kant's call to think for oneself is also a question of information design.
IV.4Personal knowledge needs personal control. A record of your own life or health is only useful if it never has to leave your hands.
How reliable can search be across the legal texts of a country with four official languages? And when does a correct answer become a liability?
Grounded insemantic access to federal legislation through linked open data (in trial); learning corpora on Swiss taxes, insurance and law.
VSpeech and Language Technology
Common threadSpoken words turned into text that can be trusted.
Walter Ong showed that speech is not imperfect writing but a different way of thinking. A transcript is therefore always an interpretation.
V.1Speech is the least structured kind of data, and the most revealing. Handling it requires honesty about uncertainty, not just a recognition algorithm.
V.2No raw transcript is published. Between recognition and publication there is always a review: correction, a sceptical second reading and clear marking of whatever remains uncertain.
V.3People mix languages, so the system must too. Real speakers switch languages within a sentence, and a useful speech system handles that.
V.4Privacy by staying local. Other people's speech is processed where it was spoken. The audio never leaves the device.
V.5Grammar can be explained. Even the famously arbitrary gender of German nouns largely follows rules that can be weighted, tested and taught.
Can we give feedback on spoken mistakes without changing the way people speak while they are being observed?
Grounded ina multilingual speech pipeline with mandatory review; live analysis of speech during meetings.
VIGeoinformatics and Mobile Computing
Common threadTerrain made readable from real tracks, not guesses.
Michel de Certeau compared walking to speaking: every route says something about the person who took it.
VI.1The public map is not the road you rode. Confidence should be visible: certain, probable or speculative. It comes from your own tracks, not from anonymous votes.
VI.2Everything is a file. A publishing chain with no writable server has almost nothing to attack and almost nothing to run.
VI.3Navigation is a function of position and intention. Its logic can be tested without the map on the screen.
VI.4Unknown terrain is revealed step by step, and only from openly licensed data. The licence is part of the data, not a footnote.
Can a community map live on without any server? Or does every shared resource eventually need a keeper?
Grounded inan offline-first navigator and recorder for e-bike routes; arrival guidance to remote alpine chalets.
VIIVisual Computing, Media Informatics and Digital Heritage
Common threadSacred and cultural spaces that can be visited from anywhere, faithfully.
Benjamin feared that reproduction would destroy the aura of a work of art. We believe the aura of a place can survive, if what we reproduce is not an image but a path.
VII.1A 360° virtual tour of a sacred space must keep the approach, the sequence and the gaze, not just the surfaces.
VII.2One grammar of movement. An automatic guided tour and a visitor's own gesture use the same mechanism: a timeline of simple steps.
VII.3Perception is measured, not guessed. Speed, zoom and framing are tuned by analysing the running tour frame by frame, then fixed as testable values. What used to be a craftsman's instinct, Polanyi's "tacit knowledge", becomes explicit.
VII.4Scale needs positions as data. The catalogue of objects is kept separate from the scenes in which they appear.
At what size does a hand-tuned experience have to become a system? And how much of its aura survives the change?
Grounded ina 360° virtual tour of an Orthodox church, with guided walk, object catalogue and chronology; a bilingual commented edition of books of the Bible.
VIIIHuman–Computer Interaction and Experience Engineering
Common threadInterfaces that people understand without effort.
Heidegger observed that a good tool disappears in use: we notice it only when it fails. Good interaction design is the art of that disappearance.
VIII.1Frustration can be measured. Interfaces are audited against a written standard, on explicit criteria, not against taste. No finding is dismissed as minor.
VIII.2The sixty-second test. A usage analysis is complete only when we can describe, without gaps, what the user does in the next sixty seconds.
VIII.3More power only when needed. The user gives one consent and meets no jargon and no visible machinery. Heavier tools appear only when the situation calls for them.
VIII.4Premium means restraint. Quality shows in what an interface chooses not to do.
Can perceived quality be measured without reducing it to numbers that miss exactly what people perceive?
Grounded ina family of audit tools for frustration-free and premium interfaces; the brand architecture of a product family.
IXDistributed Systems, Networks and Security
Common threadSecurity that can be explained, not just claimed.
In 1883 Kerckhoffs required that a system stay secure even if its design is public. Our systems follow that rule. This page, on purpose, still does not describe them.
IX.1Sovereignty is a question of topology. Where data lives, who can reach it and which law applies are architecture decisions, not contract clauses.
IX.2Explain and justify. Every diagnostic finding comes with an explanation and a measured reason. Automatic fixes are limited to what can be undone, and every change is rehearsed first.
IX.3A protective shell has two sides. It must keep attackers out and stay clear for the administrator inside. Every serious finding is checked by someone who tries to disprove it.
IX.4Nothing is deleted before its copy has been verified.
In networks run by non-specialists, how far should automatic repair go without supervision?
Grounded invendor-neutral administration of mixed network devices, including undocumented ones (in trial); a security audit of an administration layer.
XTechnology-Enhanced Learning and Simulation
Common threadSkills trained by practice, measured and replayed.
Vygotsky located learning between what we can do alone and what we can do with help. Learning tools are built for that space.
X.1Competence is a map, not a list. Knowledge-space theory (Doignon and Falmagne) orders a subject by what must be known first. Questions are then generated to fit each learner.
X.2Professional judgement is trained like chess. Each situation is a position and each answer a move. Moves can be replayed, varied and compared.
X.3Deliberate practice closes the loop: plan, perform, review. The gap between plan and performance becomes the material for the next session.
X.4Several critics see more than one. Feedback split into distinct voices (structure, challenge, language) is more useful than a single verdict.
Does replaying situations like games deepen judgement in facilitation and mediation? Or does it only teach the game?
Grounded ina deliberate-practice environment for public speaking, in regular use; a simulator for Scrum Masters in large agile organisations (in trial).
XICollaborative Systems and the Remote Office
Common threadClients and teams working on the same documents, wherever they are.
In 1967 Melvin Conway observed that systems mirror the way their makers communicate. We therefore design the communication first.
XI.1The client is a co-author, not a recipient. Client and team write, comment and annotate the same documents in one shared space. Every remark stays attached to what it is about, instead of getting lost in email.
XI.2Presence is information. Seeing who is working on the same document right now avoids collisions and waiting.
XI.3The office travels with the document, not with the building. Editing, calling and note-taking happen in the same system where the work lives. Because it is self-hosted, the documents stay in Switzerland.
XI.4A meeting ends with a document, not with a memory. The minutes are written straight into the system of record, where decisions become tasks.
XI.5One portal, many projects. A new client joins by configuration, not by copying code.
How much of a team's unspoken coordination (a glance across the desk, a quick question) can a remote office reproduce without turning it into surveillance?
Grounded ina client collaboration portal used across several mandates, with documents, spreadsheets, annotated mock-ups and live presence; a self-hosted office suite with document editing, internal calls and meeting minutes integrated into an ERP.
XIICommunity Informatics, Culture and Public-Interest Technology
Common threadTechnology for communities, culture and people with fewer means, built with the same care as for paying clients.
Dostoevsky wrote that beauty will save the world. We do not claim that much. We only refuse to give a community less beauty or less rigour because it has less money.
XII.1A community must be able to keep what we build. Parishes, clubs and associations rarely have an IT budget. What we give them must run at almost no cost and outlast the volunteer who started it.
XII.2Taking part must cost seconds, not vocabulary. A sign and a photo are enough to report what one sees. A universal language lets people take part whatever their education, their mother tongue or the age of their phone.
XII.3Explain in the reader's language. People new to a country meet taxes, insurance and law in a language they are still learning. Official terms therefore stand next to an explanation in their own language.
XII.4Sacred and artistic works deserve the same craft as commercial ones. A church, an icon, a chant or an exhibition is documented as precisely as a product catalogue, and with more respect.
XII.5A community's memory is an archive. Photos, films, sermons and chants outlive those who recorded them only if someone puts them in order.
XII.6A network that observes can also coordinate. The same simple reports that show the weather can show where help is needed. People, resources and training can then be sent where they are needed, when they are needed.
How can technology serve those who cannot pay, without becoming a favour that depends on one person's goodwill?
Grounded inthe digital presence of Orthodox parishes in Switzerland and Romania, with a 360° virtual tour and a chant-learning module; a bilingual commented edition of the Bible; the digital and printed apparatus of an art exhibition; tools for a volunteer public-speaking club; learning material for Romanian speakers in Switzerland; a free navigator for cycling routes, with ride stories and live encouragement shared by link, without accounts; a participatory weather-observation network (suspended).
XIIIHealth Informatics and the Cartography of the Body
Common threadA person's own medical findings, turned into a map they can understand.
In 1543 Vesalius drew the human body so precisely that anatomy could be learned from pictures. We bring the same care to one person's own findings.
XIII.1The body can be read like a map. The findings of an MRI report are placed on a three-dimensional anatomical atlas. A sentence in a report becomes a visible place on the spine.
XIII.2Load is calculated, not guessed. A musculoskeletal model with more than two hundred muscles estimates how much a posture or an exercise loads the spine.
XIII.3Every exercise needs a reason. An exercise is recommended only with its mechanism, the evidence behind it and the load it creates, and it is confirmed with a physiotherapist or doctor.
XIII.4Understanding reduces fear. A person who can see and name what happens in their own body talks to their doctor as a partner. We explain; we do not diagnose.
When a new scan arrives, can it be compared with the old one segment by segment, in three dimensions, without the comparison turning into a diagnosis that only a doctor may make?
Grounded ina personal knowledge system for the lower spine, with MRI findings placed on a 3D atlas, spinal load from a musculoskeletal model and exercises with mechanism and evidence (in trial, private by design); tools for volumetric reconstruction from medical image stacks.
XIVSoftware Engineering Methodology
Common threadHow we work: a clear goal, one owner per task, nothing called done without proof.
Brooks showed that adding people to a late project makes it later. We go further: a team member is justified not by the work they add, but by what only they can see.
XIV.1The goal comes before the team. It is written down, versioned and chosen from at least two different readings. Agreeing to a single formulation is not a real choice.
XIV.2Another person is justified by a new angle, not by extra hands. They must be able to see a blind spot that no tool can reveal.
XIV.3One document, one author, one place. The state of things is checked at the source, at the moment of acting.
XIV.4"Done" must be proven, not declared. The author is a witness, never the judge. There is no complete audit, only audits with a declared scope.
XIV.5Conventions are capital. A written rule, catalogued and passed on, is as valuable as a software library.
When does a larger team bring more insight than coordination costs? And how does a method built around one decision-maker cope with several who legitimately disagree?
Grounded ina register of codified practices used across more than fifty projects; a catalogue of working maxims, each traced back to the incident that made it necessary.
Research Agenda
These questions are formulated but not yet answered. We publish them as a programme, not as a portfolio.
- Self-sovereign identity: verified once, proven many times, with selective disclosure anchored in a national eID.
- A zero-knowledge personal vault as the layer of trust beneath identity and contracts.
- Contracts that connect to legal enforcement, from the first reminder to Swiss debt collection.
- Qualified electronic signatures under ZertES and eIDAS for regulated small businesses.
- Payments recorded without holding funds, including settlement in stable digital currencies.
- Measures of project progress that do not change when the work is split differently.
- Recognising finger-spelling from hand geometry as an accessibility tool.
- Ad hoc deployment from shared observation: community reports that show where staff, volunteers, resources and training are needed, and allocate them on the spot.
On Discretion
This page states theses and open questions. Methods, architectures and tools are discussed in person and under mandate. The Clear Team takes on a mandate where one of these questions is truly at stake, and says so when none is. If you recognise your own difficulty in one of them, tell us.