Newlife

newlife

[](https://pypi.org/project/newlife/) [](https://github.com/hzshen88/newlife/actions/workflows/lint.yml) [](LICENSE)

**Turn a curiosity into a question that can be settled, freeze what would count as an answer before running anything, and leave a verdict that can be audited from git alone.**

You talk to your AI; it runs the commands. Setup is two lines:

python -m pip install newlife
newlife start my-research

Open `my-research` in Claude Code, Codex, or any AI tool that reads `SKILL.md` files, and say:

Explore this with me: *your curiosity*

`NEXT.md` in that folder says what happens from here. If you get lost, ask your AI "how does this work" or "which step are we at"; the installed `newlife` skill answers in your terms. [`docs/first-question.md`](docs/first-question.md) walks one question end to end in about ten minutes, with the real output of every command.

Requires Python 3.12 or newer and `git` (the freeze is a git commit; that is what makes the timestamp real). `start` tells you if git has no identity to sign with, rather than guessing one.

What happens, in your words

1. Explore. Your AI asks one or two honing questions, then walks one edge at a time and keeps a map silently. You answer and decide. (This stage is the exloop skill from the `exloop` package; newlife depends on it, so the same install brings it.)
2. Hand off. When a boundary can be stated as "what measurement would make the answer different", the AI says so once. If you agree, a question folder appears with the exploration record inside it.
3. Is it worth asking? The AI asks only what the map cannot supply: who would bet the other way, is the answer settled by design or by running, who changes wha