Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

werr

Experimental Jev wire-compatible engine that claims to produce typed decisions from Mandelbrot fractal dynamics rather than a trained model.

Serviceresearch
View on GitHub ↗Live demo ↗

Overview

werr describes itself as a zero-memory System-One decision engine that is wire-compatible with TypeSafe Jev. Instead of running a trained model, it says it synthesizes noul, choice and score decisions from deterministic Mandelbrot fractal escape dynamics and quadrant subdivision, so no weights are held in memory. It offers an in-browser simulator and benchmark arena, and claims top placement on JevBench.

The README is dominated by badges, preprints, archive records and links to unrelated projects (formal verification, physics and biology monographs), and its framing is unusual. Readers should treat its benchmark claims with strong skepticism and verify independently; the README itself links a benchmark integrity report noting metric shifts between JevBench versions. It is included because it targets Jev's API, not as a recommended decision engine.

What the decision model decides

TypeQuestionAnswers
OtherNoul, Choice and Score answers on the /v1/systemone wire format—

How it works

  1. Send a Jev-format request to werr.
  2. The engine maps it to fractal computations.
  3. Typed answers are returned in Jev's format.

Strengths

  • Runs on CPU with no model weights.
  • In-browser simulator.

Caveats

  • The decision method is unconventional and its claims are unverified.
  • README mixes unrelated research and promotional material.
  • Benchmark results should be independently checked.

Getting started

See the live simulator at https://pcworm.github.io/werr/

At a glance

Type
Service
Status
research
Language
Python
Runtime
CPU engine, browser simulator
Has code
yes
Has tests
yes

Use cases and integrations

How curated lists describe it

  • Zero-memory System-1 decision engine & TypeSafe Jev wire-compatible runtime…

    wh000wh000/awesome-jev-live

  • Zero-memory System-1 decision engine & TypeSafe Jev wire-compatible runtime powered by Mandelbrot wave dynamics (JevBench #1).

    hellogumbo/awesome-jev

Frequently asked questions

Is werr a trained model?

No, it claims to derive decisions from fractal dynamics rather than learned weights.

Should I trust its JevBench claims?

Verify independently; the claims are unusual and self-reported.

Is it compatible with Jev clients?

It describes itself as wire-compatible with /v1/systemone.

Related projects

  • systemone-lite

    Toy local Jev-shaped decision API on a 0.5B model that answers yes/no, choice and score questions by scoring option tokens.

    ServiceprototypePython
  • jiwo

    Small open decision models (0.8B and 4B, from Qwen3.5) with a server that speaks Jev's /v1/systemone format in one forward pass.

    ServicebetaPython
  • minicpm5-jev

    Serves MiniCPM5-2B-8bit on Apple Silicon as a Jev-compatible /v1/systemone service and reproduces its JevBench score.

    ModelprototypePython
  • reflex★ 166

    Small open decision model on a frozen Qwen3.5-4B serving /v1/systemone in about 200 ms, benchmarked against Jev on JevBench.

    ServiceresearchPython
  • Carve-Jev

    Open Apache-2.0 Jev-like model family on Qwen3.8-27B that reads one-token candidate preferences and serves a TypeSafe-compatible API.

    ServiceresearchPython
  • minicpm5-2b-jev

    Open MiniCPM5-2B decision model serving /v1/systemone: typed choice, noul and score probabilities in one forward pass.

    ModelresearchPython