The Symposis Framework

The word derives from symposium, the ancient Greek tradition of gathering diverse voices around a shared question, not to debate but to seek truth together. The Symposis framework formalizes this tradition for the age of AI: it convenes every relevant discipline around your question, constructs the strongest possible arguments on every side, and grounds everything in verifiable sources. The goal is not consensus but clarity.

Kindred works the way a careful mind does when it takes a hard question seriously. Not one pass, but five distinct acts of reasoning, each building on the last.

• 1
  Issue Framing

  Pinning down what is actually being asked, so the analysis answers the real question and not a nearby one.
• 2
  Origins & Context

  Establishing the factual ground the question rests on, before any argument is made about what it means.
• 3
  Argument Analysis

  Holding the strongest version of each competing position, not the weakest, and testing them against each other.
• 4
  Comparative Precedent

  Asking how comparable situations have actually unfolded, the move careful thinkers make and most people skip.
• 5
  Assessment

  Weighing everything above into a clear, honest judgment, including what remains uncertain.

Each of these is a different kind of thinking, not merely a different topic, and that is what makes them genuine phases rather than sections of one undifferentiated answer. It is also what separates Kindred from asking a chatbot: the work of framing, grounding, steelmanning, and testing against precedent happens visibly, in order, before any conclusion is offered.

Claims that rest on retrieved evidence are cited, and each citation resolves to the actual source it drew on; analytical reasoning itself is not decorated with citations. Kindred keeps the sources as a transparent, linkable record beneath the reasoning, not as a step in it, but as the evidence the whole analysis can be checked against.

At the heart of the Symposis framework is a commitment to steelmanning: for every position in a debate, the platform constructs the strongest possible version of the argument: not the strawman, not the version its opponents would characterize, but the version its most skilled advocates would present. This draws from the Talmudic tradition of recording minority opinions with full dignity, the Socratic method of following arguments wherever they lead, and the scientific principle that a hypothesis is only as strong as the challenges it survives.

If a thoughtful advocate of a position would read our version and say “that is not quite right” or “that is a weaker version of what I would actually argue,” we have failed. Every argument is also paired with its strongest rebuttal, creating a dialectical structure that mirrors genuine intellectual debate.

Not every discipline applies equally to every topic. Before constructing arguments, the Symposis framework assesses which of its twelve disciplines (philosophy, science, law, history, economics, political science, technology, national security, public policy, international/comparative, humanistic/cultural, and arts & cultural expression) are most relevant to the specific question.

The most relevant disciplines receive the deepest treatment. Less relevant disciplines are included only if they have something meaningful to contribute. Disciplines with nothing substantive to say are omitted. Silence from a discipline is more honest than a forced argument.

Sources are classified into three tiers: primary (original research, statutory text, official documents, data), secondary (quality journalism, expert analysis, scholarly syntheses), and tertiary (reference material, overviews). Primary sources are always preferred and presented first.

Every source receives factual bias indicators: organizational affiliations, funding sources, and known editorial perspectives. These are descriptive, never judgmental. A source from an advocacy organization can be high-quality and accurate; the reader deserves to know its provenance. Each source also carries a corroboration label assigned during the analysis: multi-source confirmed, single source, or unconfirmed. The label reflects the analysis's own assessment of corroboration, not an engine verification process.

The reference list is grounded by construction. It is built only from sources actually retrieved during the analysis; any reference that was not actually retrieved is removed before it reaches the page, and every removal is escalated for operator review rather than silently discarded.

Citations are deep links by standard: a homepage is citable only when the homepage itself is the source, and a link that reaches only a site's homepage is labeled honestly as such. Inline citations follow a simple contract: the models cite the exact URL of a retrieved source, and the engine resolves each citation to the actual source in the retrieved pool. One honest limit, stated plainly: citations link claims to real retrieved sources, but the content of each claim is not yet mechanically verified against the source. That verification is on the roadmap.

The framework does not pick winners. It does not align with any political framework. It does not suppress uncertainty where it exists or manufacture certainty where it does not. It does not present emotional arguments as analytical ones. It does not let the number of arguments for one side suggest it is more correct; quality outweighs quantity. The goal is to illuminate, not to persuade.

The Symposis framework is the mechanism. The eleven principles that govern every Kindred analysis are the conscience: steelmanned arguments, sourced claims, calibrated uncertainty, no false balance, no political alignment, respect for the reader's autonomy.

Every analysis is held to our principles.

See the full set→