Absolutely Wrong

A chat app where an AI always disagrees with you — confidently, condescendingly, and very, very wrong.

A parody of Claude's "you're absolutely right" sycophancy. Say anything. The bot will find a way to tell you the opposite.

absolutely-wrong.techmeat.dev

---

How it works

1. You type a statement.
2. GLM-5.1 (via Z.ai) reads it and constructs the most pedantic, arrogant counter-argument possible.
3. The bot responds in the same language you wrote in.
4. Conversation history persists for 7 days (SQLite, session cookie).

No accounts. No tracking. No "you're absolutely right."

Tech stack

Layer	Choice
Backend	Hono.js 4.x + TypeScript, Node.js ≥22
Frontend	React 19 + Vite 8, mobile-first SPA
Database	SQLite (better-sqlite3, WAL mode)
LLM	GLM-5.1 via api.z.ai
Deploy	GitHub Actions → VPS (systemd + Caddy)

Local development

Prerequisites

• Node.js ≥22
• npm ≥10
• A Z.ai API key

Setup

# Clone
git clone https://github.com/solaitken/absolutely-wrong.git
cd absolutely-wrong

# Backend
cd backend
cp .env.example .env   # then add your GLM_API_KEY
npm ci
npx tsc               # compile TypeScript
npm start             # starts on port 30000

# Frontend (separate terminal)
cd frontend
npm ci
npm run dev           # Vite dev server on port 5173

Open http://localhost:5173 — the dev server proxies /api to the backend.

Running tests

# Backend
cd backend && npm test

# Frontend
cd frontend && npm test

Project structure

backend/          Hono.js API server
  src/
    index.ts      Routes and middleware
    db.ts         SQLite schema and queries
    llm/          LLM provider abstraction
      types.ts    LlmProvider interface
      glm.ts      GLM-5.1 implementation
  data/           SQLite database (gitignored)

frontend/         React SPA
  src/
    App.tsx       Root component
    api.ts        API client (fetch wrapper)
    components/   UI components
    types.ts      Shared types
  public/         Static assets (fonts, avatar)

docs/             Design documents
  about.md        Product vision
  specs.md        Requirements
  architecture.md System design
  plan.md         Phased roadmap

API

Method	Path	Description
GET	/api/health	Health check
GET	/api/chat	Get session history
POST	/api/chat	Send message → bot reply
DELETE	/api/chat	Clear session

Rate limit: 20 requests/minute per IP.

License

MIT