One question, twenty-two minds. A local app I built that routes any decision through a ring of historical thinkers, presidents, CEOs, and three sacred books — each a distinct reasoning lens — then has an oracle named Xiao read the room and deliver a single, spoken verdict.
Michael Paycer
Most advice tools collapse toward one confident answer. This one does the opposite — it forces a question through Stoic discipline, practical execution, moral burden, scientific proof, strategy, mindfulness, and contrarian interrogation at once, so the blind spot of any single lens gets caught by another.
14 permanent advisors, always convened
8 optional specialists
3 sacred texts — a moral check
Xiao synthesizes one verdict
Python · Streamlit · Claude API
Wake word “Xiao” · ElevenLabs
These sit in a circle around Xiao and answer every question. Each is defined by the question it always asks.
Called in from the sidebar when a question needs their narrow strength.
A separate moral mirror that runs after the council speaks and before the verdict. It does not make the decision smaller; it makes it cleaner.
An empty chair — not an advisor. It asks the one thing analysis forgets: who pays the price if this is wrong?
Records the decision, the assumptions, and later what actually happened — so the council becomes a record of your judgment over time.
Learning and studying. Feynman forces me to explain a topic simply or admit I can't; Curie separates what I know from what I'm assuming; Socrates keeps asking the question under the question until the gaps show.
Ethics. Kant tests whether a rule survives being universalized; Lincoln weighs the human cost when every option hurts; the Conscience Seat checks whether I'm seeking the right answer or protecting the version of myself that doesn't want to change.
Philosophy and theory. Run a genuine question and watch the frameworks diverge — Stoic control, Taoist non-force, Socratic interrogation, Kantian duty. The disagreement is the point.
Friends, family, and everyday calls. Relationship questions route to Thich Nhat Hanh and the Dalai Lama; the Affected Person chair keeps asking who actually bears the cost of the choice.
“Take the role only if you can name the specific thing it buys that your practice can't, and say it out loud without flinching. If you can't, the offer is a test of nerve, not a better life.”
Every advisor is one authored block of text — the framework field in the code. I wrote them by hand to a single template so 22 very different figures produce output that is distinct in voice but comparable in shape: identity, the seat and lens they hold, the question they always ask, what they're best and worst for, their blind spot, how they disagree, and a fixed output format.
At runtime a prompt builder assembles the active advisors into one request and requires strict JSON — core issue, per-advisor verdicts, the disagreement, a recommendation, the hard question, and Xiao's spoken verdict. Adding or revising an advisor is just editing one well-shaped block of prose.
A local app that sends one question to 22 advisors, each a distinct decision framework modeled on a historical thinker, then has an AI oracle named Xiao synthesize the core issue, the disagreement, a recommendation, the hardest unasked question, and a final spoken verdict.
The Decision Council is broader and built to decide: it mixes philosophers with presidents, CEOs, the field observer Jane Goodall, and three sacred books. The Council of Philosophers is philosophers only and built to understand. Both share Xiao, voice, court roles, and the decision log.
It runs locally in Python and Streamlit against the Anthropic Claude API, with optional ElevenLabs voice and browser speech input. Each advisor is one authored framework block assembled into a single strict-JSON request.
Michael Paycer, a SQL Server DBA and developer in St. Cloud, Minnesota, built it as a personal project.
The Council of Philosophers runs the same engine with twelve philosophers and nothing else — built to understand an idea rather than decide.