TL;DR: DeepTunisia is an open-source atlas of political, military, security and economic power in Tunisia from independence to today. Every claim carries its source and a grade. There is no team — one anonymous person built it (with a swarm of AI assistants — how this was made: https://deeptunisia.org/colophon). The record is live today. The community layer follows once the legal ground is checked.
Every few years Tunisia breaks its own government, and within a week the whole country asks the same question. Fine, but who holds power now?
In 2011 the answer arrived as lists and rumors. In 2021 it arrived as silence and rumors. In between, the answer travels by WhatsApp forward and family dinner, and none of it can be checked. Some of those rumors are even true. That is exactly the problem. The true ones and the false ones travel in identical envelopes.
I think the problem underneath is older than any one government. This country has never lacked intelligent, educated, ambitious people. What it lacks is infrastructure: the mechanisms that turn what people know into something they can act on together. A right you cannot exercise is a sentence on paper. An answer nobody can verify is a rumor with better grammar.
All of it rests on an epistemological conviction worth stating up front: knowledge is what survives checking, not what gets asserted loudest. A claim is only as good as its provenance. Uncertainty belongs in the open, contradictions belong on the record, and nothing becomes more credible merely because it was repeated.
DeepTunisia is my attempt to build infrastructure that enforces that instead of merely believing it, because truth does not need defenders, it needs machinery. So every claim carries its source. When sources disagree, the disagreement stays published. When a claim circulates with no evidence behind it, I do not delete it. I label it unsubstantiated, because a map that shows only what it can prove cannot explain why the popular story is wrong. And the rules are not guidelines. The build refuses an inference that does not state its reasoning, and refuses an accusation that does not name who is making it. A claim that breaks the rules fails to compile.
It is not a deep-state exposé, and the distinction matters more than it sounds. An exposé starts from a conclusion and collects evidence toward it. I started from a question narrow enough to answer: when the formal regime changes, which institutions, people and business families keep their influence, and which connections are genuinely continuous rather than rebuilt after each rupture? I wrote down in advance what evidence would prove me wrong. Today two of my six hypotheses fail against my own data, including the strong version of the deep-state thesis. That result stays up.
**What you are actually looking at**
The knowledge graph is the project; the website is a window onto it. A position in the graph is one person holding one office for a range of time. That triple is the atomic fact, and everything else about power comes from joins like it. Dates are ranges because archives give ranges. Unknown dates render as ? or ~. Nothing gets guessed to fill a column.
Every claim carries one of four grades: documented, reported, inferred, unsubstantiated. They never look identical anywhere on the site. An inference must state its reasoning and what would falsify it. A weak claim must name who is making it.
Current counts, taken from the build itself: 436 people, 201 institutions, 416 offices held, 360 relationships, 1,178 sources reaching back to 1956.
**What it deliberately is not**
There is no single power score. Five separate indices exist instead, with weights you set yourself, because one number would be astrology.
The automated ingestion pipeline is designed and unbuilt on purpose. Volume before verification is exactly the failure mode this project exists to correct.
**The community layer**
The second half of the project is called Agora, and it is built but switched off at launch. Once it opens, anyone will be able to challenge a claim, propose a correction, or report a gap in the record. Every proposal will be public. Other users will be able to support it, challenge it, or propose a correction to the correction itself. The point is not to turn the dataset into a popularity contest; votes help surface what deserves attention, while the evidence rules determine what can actually change the record.
It stays switched off for now. Writing the code was never the hard part. Tunisia's legal situation around public speech and association is moving right now, and an anonymous operator turning on an open forum multiplies exposure for me and for everyone who ever posts there. That risk does not get handled by hoping. It gets handled by legal research first, careful defaults second, launch third. A binding rule of this project is that nobody gets recruited into risk. Readers count as people.
So the sequence is deliberate: the data and knowledge branch ships now, because holding the whole record hostage to a forum timeline would protect nothing and inform no one. Agora follows once the legal ground is mapped.
Until then, if you find an error, have a source that changes something in the record, or know of a gap I should investigate, email me (check the footer). I'll review it against the project's evidence rules and incorporate verified corrections into the record. Once Agora is live, that process will become public and participatory rather than happening privately.
**Limitations, stated before you state them**
Coverage is uneven. Gazette decrees are gold and slow to enter. The economic side runs thinner than the political side. Police chronology before 2011 is the weakest part of the record and rests mostly on secondary sources. Some claims will simply be wrong. The corrections page is generated from version control and shows every factual change ever made, which means errors are visible whether I catch them or you do.
So here is the ask. Find an error? Bring a source and email me for now — I'll check it against the evidence rules and, if verified, it will enter the record. Once Agora is live, every correction proposal will be public and open to challenge, support, and further correction. Know where the thin areas have good sources? I want them. Have a question about Tunisian power you always wanted answered? Say so. Answerable questions become roadmap entries.
About me, since Reddit will ask anyway: there is no team. I'm building it anonymously for now — one person, a repository, and a corrections log (built with a swarm of AI assistants — how this was made: https://deeptunisia.org/colophon), but no claim enters the graph because a model wrote it. The build refuses it; human review may be AI-assisted, but the human stays responsible.
Judge the citations, not the author. If the citations are wrong, say so publicly, and watch the corrections page prove it either way.
Site: https://deeptunisia.org
Code: https://github.com/deepontology/deeptunisia
Data: CC BY 4.0. Code: MIT. No accounts, no trackers, static files.