Transparency

01

Funding

Core Group is privately held. At this stage we are funded by our founders and a small group of private investors who back the methodology, not its conclusions. We do not take funding from political parties, faction-aligned foundations, or actors with operational stakes in the conflicts we analyze. As we build out our analytical engagements, we will disclose any single source representing a material share of revenue.

02

Independence

Convergence is methodological. It does not adjust output to align with client preferences, political affiliations, or institutional interests. Our analysts form views from the evidence available; clients receive what the methodology produces, including findings they may not have wanted. The methodology is the contract.

03

Editorial standards

Every claim we publish carries a source grade. One axis grades the source, from A to F. The other grades the information, from 1 to 6. A reader can see, on the face of a finding, how much weight it can bear. No event is treated as established on a single stream. Convergence requires corroboration across independent sources within a fixed window of place and time before a claim is published as fact. Contested claims are published as contested, with the disagreement shown rather than settled by assertion. We also grade our own forecasts in public, including the ones we get wrong, because a method that hides its misses cannot be audited.

04

Conflicts of interest

We disclose relationships that could bear on our analysis. Where a client, funder, or partner holds a stake in a subject we cover, the relationship is declared or the engagement is declined. We do not accept work that requires a predetermined conclusion, that shields a party from scrutiny, or that times a finding to a party's advantage. An analyst who cannot hold a subject at arm's length does not carry it. These are the commitments that govern our funding and our independence, applied to the single engagement.

05

Governance

Core Group is founder-led. Editorial authority rests with our analytical leadership, and no commercial function can override a published finding. At this stage the method is the governing structure. The grading, the corroboration threshold, and the public record are the controls that hold the work to account, not a title or a committee. As the firm matures we will add formal structure, an advisory function and an independent review of our editorial standards, and we will name it here when it exists.

06

AI and knowledge production

The right tier. We treat intelligence as a set of tools, not one. A narrow model built for a single job beats a larger one pointed at it, and most machine error is the same mistake wearing different clothes: reaching for more scale than the task needs. Deterministic code decides an answer in advance. A trained model guesses from what it learned. A person, in this frame, is also an inference machine, and the distance between the person and the model is architecture and scale, not kind. So what changes when the instrument changes is not only speed. It is what can be known at all, and where the line sits between a surprise and a thing that was readable in advance. Learning to set the instrument earlier, and to know which tier a question needs, is the discipline. We are still learning it, in the open, which is what Praxis is for.

The roster and the line. The house runs on a roster of specialized agents, each narrow enough that when it is wrong a person can catch it. Some do the finding and reading across languages and sources at a scale no human team could hold. Some coordinate a production line, keep the record clean, or track access, and some now run continuously, as infrastructure rather than as a tool with a session and an end. A human owns each line of work. The analyst points the instrument, sets the thresholds it works to, reviews what comes back the way an engineer reviews a submission, and owns the meaning of every finding we publish. The machine finds and reads. The human keeps the judgment. That division of labor is Convergence, and it is the line we do not move, even as the machinery beneath it keeps changing.

The power and its limits. The reach cuts both ways, and we say so. The same power that exposes what a state hides can expose the people it should protect, and we hold it to the same independence and conflict rules as the rest of our work, bound by why a finding matters rather than that we can reach it. The roster carries a second, quieter danger. It is very good at producing volume and poor at knowing when it is wrong, and it will not volunteer that a shortcut it took quietly changed what a number downstream means. That is exactly why the human line is load-bearing and not ceremonial: a named analyst stands behind the finding because the machine cannot stand behind itself. This is a practice we are building, not a settled doctrine, and we would rather describe it honestly as it changes.

07

Rights and reuse

Our published work is copyright Core Group. We want the analysis to travel, so quotation and citation are welcome with attribution, and you are free to share our publications in full, unaltered, with the attribution intact. Republishing our work as your own, or reselling it, requires our permission. When in doubt, write to us. We would rather say yes than see the work go uncredited.