tereo 0.1.0

Proof layer for software changes. Keep only what proves itself.

TEREO

AI writes. TEREO proves.

TEREO is a proof layer for AI coding. keep only if gain > noise Each kept receipt becomes the next baseline.

It watches one small change, reruns one fixed check, and writes one readable receipt. The check can prove a bug fix, a behavior change, or a speedup.

TEREO comes from Ancient Greek τηρέω (tēréō): to watch over, guard, preserve.

One proofboard.

Demo

1. Install TEREO.

pip install git+https://github.com/kim-woojoo/tereo.git

2. Feel the loop.

tereo demo

If you are working inside this repo instead:

pip install -e .

The demo should close with something like this:

scorecard: latency: 10.000000 ms -> 8.000000 ms; 2.000000 ms better (20.00%)
verdict: keep
confidence: high

Quick Start

1. Install TEREO.

pip install git+https://github.com/kim-woojoo/tereo.git

2. Run TEREO locally.

tereo --help

3. If you prefer module execution, run:

python3 -m tereo --help

Use your own check. Keep your own workflow. Make it yours.

TEREO is small on purpose. Fork it. Keep it. Make it yours.

Public Commands

tereo demo
tereo prove
tereo show
tereo log

Everything else is there to support this surface, not to replace it.

Why This Is Hard

• one good-looking run can lie
• gains drift
• noise changes verdicts
• checks matter more than vibes
• hold is different from keep

Narrow By Design

TEREO is strongest when one fixed check can judge one small change. That is a feature, not a full replacement for broader product judgment.

Read Next

• START.md Start here.
• runtime/README.md The runtime, advanced commands, receipts, and proofboard.
• skill/README.md The portable rulebook for humans and agents.
• WHY.md Why TEREO stays this small.

If the project is easy to explain, it is easy to trust.