01 Rule-based dialogue, made visible

A conversation
you can inspect.

ELIZA Lab turns a famous interaction pattern into a small, honest learning tool. Every reply comes with its rule trace. Nothing leaves your browser.

  • No accounts
  • No transcript storage
  • Not therapy

02 Live experiment

Talk to the rulebook.
Then open it.

Use a harmless, fictional prompt. The inspector shows what matched and how pronouns were reflected. The transcript exists only in this tab and clears on reload.

LOCAL SESSION / NO STORAGE

ELIZA / 00

Hello. What is on your mind today?

Fictional text only · 512-character limit · 40 visible turns

03 Under the surface

Small mechanism.
Large illusion.

01 / NORMALIZE

Read the shape.

Whitespace and casing are normalized. The engine never sends the sentence elsewhere.

02 / MATCH

Choose one rule.

Patterns run in a documented order: safety boundary, greeting, feeling, reason, ownership, question.

03 / REFLECT

Turn the sentence.

A tiny pronoun map changes “my” to “your” and returns the thought as a question.

04 / EXPLAIN

Show the seam.

The selected rule and transformed fragment stay visible so the illusion never hides the method.

04 A clear boundary

A mirror is not a clinician.

ELIZA Lab is an educational software experiment. It does not understand a person, assess risk, provide treatment, or replace qualified care.

A small list of explicit safety phrases stops the simulation and points toward immediate human help. It is a boundary, not reliable crisis detection.

Read the safety and privacy model

05 Refactored with intent

The useful idea survived.
The unsafe product did not.

The original 2023 Telegram prototype stored conversations and implied therapeutic use. This edition removes Telegram, MySQL, OpenAI and the PHP admin panel. What remains is a testable Rust engine and a transparent browser demonstration.