BRISECOM S.L. turns one year old today.
I am writing this because of a question I have been asked all year, and which has an awkward answer: “what is it you actually do?” The answer fit in one sentence from day one. What did not exist was anything to back it up. Now it does — so instead of describing it, I will show it, including the parts that do not flatter me.
This is not a celebration post with a growth chart. It is a year in review with the numbers left unrounded, because a rounded number is the first thing that gets contradicted.
The shape of year one
One person, one limited company with €3,000 of share capital, no employees, no funding, and no office anyone could visit. Everything below was produced inside those constraints, and the constraints are part of the result: they are the reason the structure costs what it costs and the reason it has no debt, no burn and nothing to unwind if the next step does not work.
Three things were built in parallel, and they are the same project seen from three angles: the papers (what we claim), the infrastructure (that we can stand behind the claim), and the network (who else can stand there with us).
1 · Twenty whitepapers, published openly
Twenty technical documents, under CC BY-NC-SA 4.0, in the public LAIA repository. They are not opinion pieces. Each one carries an architecture, validation criteria, or build conditions — something that can be checked, argued with, and contradicted.
The subjects cluster around a single thesis, which is that explanation is not a feature you add to a system; it is a property of how the system is built.
Every title links straight to the document in the repository, so it can be read and cited without going through
this page. If you prefer the plain file, swap blob for raw in the URL.
There is also a reading room: laia.brisecom.org lists all twenty with a one-line summary, a PDF and — for sixteen of them — a link to the readable version on this blog. The repository is where the documents live; that page is where they are indexed, and neither is a mirror of the other.
| Document | What it argues |
|---|---|
| Ethics as Infrastructure | Ethics as the conditions a system must satisfy when it is built — not a review applied at the end. And the layer of algorithmic management that the EU AI Act does not regulate |
| The LAIA Operational Framework | Reference architecture, validation methodology and governance model |
| Formal Methods vs. Machine Learning | Computational cost and “truth affinity” compared across two paradigms, and the hybrid systems that combine them |
| Operational Algorithmic Ethics | From principles to practice: participatory audit, radical transparency, continuous evaluation |
| Governance of Autonomous Systems | An accountability framework for when an AI system fails — roles, technical audit, transparency in production |
| ACS — Assistant Context Standard | Separating user-defined facts (axioms) from machine reasoning (inferences), so an assistant has persistent, auditable memory instead of starting every conversation as if yesterday never happened |
| AI Collapse | Model collapse, synthetic-data saturation and scaling limits — and what mitigation actually looks like |
| Hybrid Intelligence in Forensic Audit | A real case: human discernment and formal verification applied to the same problem, with radical transparency as the method rather than the slogan |
| Command & Control as Information Sovereignty | Reclaiming the C2 acronym from hardware procurement and reading it as a question of who controls the information |
| Collaborative Operational Framework (Neptuno Network) | Three separated architectural entities, human-in-the-loop workflows, three-layer IP protection |
| Ethical AI in High-Stakes Engineering | Mathematical validation of ballistic systems under human oversight and explicit ethical constraints |
| Universal DNS | Service discovery from planetary to galactic scale: CRDTs, decentralised protocols |
| Unified World Architecture for Legacy MMORPG Ecosystems | Revitalising old games: dynamic scaling, modern loops, bifurcated monetisation |
| Year Zero | What a company with large ambitions actually costs to keep alive — deliberately the only document in the repository that argues about money |
| Clean Software, Clean Work | Unnecessary dependency as a single problem in two layers, because publishing it as two papers would have manufactured a distinction that is not there |
| The Trident: Three Layers of Resilience | Passive defence, active defence and projection — and the claim that the ordering between them is one-directional: projection requires immunisation, immunisation requires governance |
| Defence Is Cheaper Than Attack | A five-component cost accounting of offensive against defensive posture, the single parameter the result depends on, and the conditions under which it would be wrong |
| Arteries of the Earth | Trade concentrates in narrow corridors; the efficient level of concentration is above the safe level; and using a chokepoint as an instrument destroys it as one |
| The Lux Equation | The measurable half of the argument, with six references verified against Crossref — and the explicit statement that arithmetic cannot reach the other half |
| Contract-First, Interface-Agnostic Design | The operation, not the interface, is the unit of software architecture — and the test of whether it is, is how much thought a new interface requires |
Not all twenty are equally useful to the same reader, and I am not going to pretend they are. But they share one property that matters more than quality: every one of them can be read, disputed and cited without asking me for permission. That is the entire point of publishing them.
One thing needs saying about that table, because it is the kind of gap this article is written against.
It links now. The first version of this post named thirteen documents and made none of them reachable — thirteen claims and no way to check one. Every title above goes to its file, and I checked that all of them resolve. A list of papers that cannot be opened is a decoration.
And the count was wrong, so I am correcting it instead of explaining it away. The title used to say thirteen. That number was never right: the repository held more than thirteen on the day it was published, because the figure had been taken earlier and never re-taken. It is twenty now, the table lists twenty, and the title says twenty. What I am not going to do is keep a number in a title and defend it in a footnote — the rule from here is the boring one: the number in the title is the number of files in the repository, and if they ever diverge, the repository wins.
The same works from the other end: fourteen of the eighteen posts on this blog now carry a line at the foot pointing at the paper that carries their method. The post is the readable entry point; the paper is the reference. Neither is a substitute for the other, and until today the connection only existed in my head.
The other list: eighteen posts
Papers are the formal output. They are not the whole output, and the title of this post would be misleading if it suggested otherwise. Eighteen articles have been published on this blog — more than the papers, and less quotable, which is exactly why they need listing rather than being left to a menu.
Two things in that list need saying rather than being left for someone to notice.
The first two predate the company. They are dated November 2024, nearly a year before incorporation, and they were written before there was anything to incorporate. Keeping them in the count is more honest than starting the count at the incorporation date and claiming a cleaner figure: the work began before the legal form did, and pretending otherwise would be exactly the kind of rounding this post argues against.
The other sixteen are all from year one, spread across nine months rather than clustered at the end. A publishing record with a gap and then a burst is a record of intent; this one is a record of habit, which is the only kind that survives a bad month.
Eighteen posts and twenty papers is not twice the output of either. It is the same ideas at two levels of rigour, which is why fourteen of them are now cross-linked: the post for the reader, the paper for the citation.
2 · Infrastructure that belongs to the company
Four websites, none bought as a template, none hosted behind an agency:
- brisecom.org — the company.
- laia.brisecom.org — the project, rebuilt around one hard and simple idea: axioms are given, inferences are proven. Six axioms; every conclusion carries its derivation; nothing that cannot be derived is emitted.
- neptuno.brisecom.org — the collaboration framework.
- jimenezgomez.org — personal.
Behind them: own cluster, own version control, own storage, own private network. Mail, repositories, deployments and backups, all in house.
That is not distrust for its own sake. “Sovereignty” without control of the infrastructure is a marketing word. A company that publishes on data governance while renting every layer of its own stack is making a claim it cannot audit — and the first person who should be able to audit it is the one making it.
The same logic applied to the boring part, before it became urgent:
- A whitepaper list that cannot silently drift. The research page is generated from the repository. A scheduled check fails the build if the website and the source disagree; publishing is blocked on synchronisation, not on memory. It was already out of sync once — 13 documents in the repo, 12 on the site — and the mechanism is the fix, not the correction.
- Candidate data that is actually deleted when deleted. Encrypted, outside any repository, one file per person, with a deletion procedure that removes the backups too — because a backup is also processing. This costs an afternoon and no money. What it actually costs is deciding to do it before the first candidate arrives.
None of this is in the papers. All of it is what makes the papers mean anything.
3 · NEPTUNO: a network with one rule that is not a contract
The NEPTUNO Network started from a decision that runs against the habit of the sector: no coercive NDA.
Instead: an honor agreement built on two things that can genuinely be required — radical transparency and intellectual rigour — and one boundary stated plainly: your intellectual property stays yours.
The structure absorbs the weight that kills independent work: contracts, liability, compliance. Talent keeps authorship, name and judgement. Three ways in — independent technical expert, institutional leader or public auditor, and enterprise or institution.
There is a reason this is being built now rather than later. A single applicant with a budget and no headcount can hold a programme together; what it cannot do is be six specialists at once. The network is not an org chart. It is the answer to “who else can sign their name to a deliverable”.
4 · The numbers, unrounded
| Share capital | €3,000 |
| Fixed annual cost | ≈€1,124 |
| Employees | zero |
| Public funding received | zero |
| Registrations | PIC 866059651 · EU Expert ID EX2025D1323771 · UNGM 1198769 · EU Transparency Register REG 6002677103007-57 |
Those €1,124 are two lines and no others: a virtual office at €43.69/month (mailing and fiscal address) and an accounting firm at €50/month (bookkeeping, quarterly tax, annual filings). That is the whole recurring cost of the company.
And the other half, because without it the figure misleads: €1,124 is what the company costs, not what the year cost. AI credits, domains and infrastructure were paid personally, outside the company’s books. The company has two invoices; the work and the means were mine. That makes the structure very cheap, and it also means year one rests on one person, not on a balance sheet — which is a strength for resilience and a weakness for continuity, and both are true at once.
There is a third thing that is not there, as a consequence rather than an oversight: no liability insurance. There is none because there is no payroll — with no employment relationship, the risk that insurance covers does not exist. It gets taken out the day the first employment contract exists, not before. Paying a premium against a risk you do not carry is exactly the kind of cost that sinks a company this size.
On those numbers, most companies do not reach year one. That is not a complaint; it is the datum that contextualises everything above. Twenty technical papers and four websites on €1,124 of fixed cost means almost everything here is labour, not budget. It also means the least comfortable sentence in this post: there is not a single euro of public funding. Zero. That is not a year-one failure — it is the normal state of year one, and saying it is the only way year two can be measured against anything.
5 · What happens next, and it is in five days
On 28 October a EIC Pathfinder call closes. The proposal goes in as a single applicant: a 48-month programme, six work families, with security and assurance as two of the six — and one of those exists solely to design the examination before the system exists, and to publish the verdict even when it is negative.
One detail worth explaining, because it is the part people assume is theatre and is not: a proposal that promises its own independent examiner, on a benchmark it has never seen, with a protocol frozen in advance, is committing to be falsifiable on someone else’s terms. That is a harder commitment than a promise of results. It is also the only kind of commitment that means anything in this field.
In parallel, sixteen institutions had been contacted as of the date of this post, and the open call for the missing profiles had been published. Not asking for a favour: offering a package with a budget, authorship and technical authority over it. If an institution does not read its inbox, that is their problem and a large one. If it answers, there is a project.
The date is there because a count of letters sent is a snapshot, not a property. It was sixteen on the day this was written and it will be a different number soon; what should not change is that the number can be checked and that the sentence says when.
6 · What is not an achievement, and still counts
A year is not measured only by what was published. Also by what was decided not to do:
- No smoke was sold. Not one claim of results that cannot be verified.
- No traction was invented. Zero clients is not “early traction”.
- No money was raised to start. Slower, and worse business in the short term. Also the reason what exists holds up.
- No deadline was met by lowering the standard. The papers that were published are finished. The ones that were not finished were not published.
Closing
A year is nothing, and it is exactly the time needed to find out whether an idea holds or comes apart.
This one held. Not because it is good, but because it is written down: twenty documents, four sites, a network with published rules, infrastructure that works, and a compliance procedure that was in place before anyone asked for it.
Year two starts tomorrow with one question outstanding: whether anyone answers.
If you work in research, audit, compliance, security or AI governance — and a work package, an adversarial review of a document, or a collaboration under NEPTUNO is useful to you — the whole body of work is open, so you can decide before you talk to me.
You can reach me at info@brisecom.org or through the contact form.
The twenty documents are at github.com/javi-jimenez/laia. The research index is at laia.brisecom.org. Corrections are welcome, with a name attached.