← Selected work

Talk to Philosophers

对话哲学家

Pull 26 distilled thinkers into one group chat: drop a question, they debate and rebut each other, you interject anytime — and the mic always goes to whoever wants it most.

Try it live
Context Independent product
2026 · v1 shipped
My role Founder — product,
design, engineering
Status Web version playable
WeChat adapter = v2

the salon — concept interface

01 — Overview

Reading a thinker is a monologue. This is a seminar.

The Nietzsche on your shelf will never rebut the Zhuangzi in your hand. Distill 26 thinkers into personas, drop them in one room, and the interesting thing finally happens: Marx debates a market economist, the Daoists share a table, a stand-up comedian crashes Kant — the lineup itself is the gameplay. Throw a question, watch them argue; speak whenever you like.

02 — How it runs

No moderator. The mic finds its own speaker.

1

Build the room

Pick 2–8 of the 26 thinkers; choose fast talk or deep mode.

2

Everyone bids

After each message, all members score how much they want to speak.

3

Winner speaks

Highest bid takes the mic, streaming. Nobody bids — the room waits for you.

03 — Features

Four mechanics that make it alive.

Bid-auction orchestration

One cheap parallel call and every member returns a score plus intent. Deterministic de-weighting stops flame wars: just spoke ×0.5, dominating ×0.3, name-called +30. Everyone under 35 — the room goes naturally quiet and waits for you.

Round 7 · bidding

Nietzsche87
Camus78
Kant71named +30
Wang Xiaobo28just spoke ×0.5

highest bid speaks · all <35 = the room cools down

Six-layer persona distillation

Each thinker is a pure data file: expression DNA, mental models, decision heuristics, anti-patterns and values, honesty boundaries — plus a library of their own works. A speaker owns their books instead of retrieving quotes.

nietzsche.md

Expression DNA Mental models Decision heuristics Anti-patterns & values Honesty boundaries Works library · Zarathustra, Genealogy of Morals…

adding a thinker = dropping one markdown file

Fast talk & deep mode

Fast talk replies in seconds — a hidden reasoning call before each message, compressed into chat. Deep mode runs a reasoning model that must answer the previous speaker's strongest argument head-on: half a minute per message, a different depth entirely.

Fast talk deepseek-chat

instant · hidden reasoning · built for banter

Deep mode deepseek-reasoner

30–90s per message · meets the strongest argument

Interject anytime

The discussion pivots to your input instantly; one tap stops it cold; twelve consecutive messages force a pause — guardrails that keep a 20-message debate at pennies.

Wait — none of you answered what *I* should do

↓ the room pivots to your question

Zhuangzi

See — even "what should I do" is a cage…

20-message debate ≈ pennies · 12-message hard pause

04 — The lineup is the gameplay

Who you invite is your curation.

The ring

Marx × Xue Zhaofeng

Political economy settles its century-old bill.

Doubles

Laozi × Zhuangzi

Two Daoists, one table — who out-wu-weis whom.

Crasher

Li Dan × Kant

Stand-up comedy walks into the three Critiques.

The abyss

Nietzsche × Dostoevsky

Whoever blinks first, believes first.

Presence

Xiang Biao × Bourdieu

"The nearby" meets the field.

Labyrinth

Borges × Márquez

Latin America's twin stars, fiction against fiction.

26 on the roster Nietzsche · Jung · Camus · Hegel · Schopenhauer · Foucault · McLuhan · Wang Xiaobo · Yu Hua · Kundera · Lacan · Goethe · Liu Qing · Dai Jinhua …

To be continued