Cascade subnet participation (SN91 / testnet 259)

Participate on Cascade by submitting a synthetic time-series generator. The subnet trains a fixed Toto2 model on your data and scores forecast quality (CRPS/MASE) — you compete on the prior, not a model you trained.

This repo implements a field-aware mixture in a private research tree, then packs a compact bland folder for Hippius.

Status: testnet 259, UID 64. Coldkey anacoluthe, hotkey cascade-gen1, Hippius anacoluthe89 (public). Deploy not yet run. Mainnet 91 later.

Quick start

1. Install upstream Cascade CLI (separate directory, Python 3.11+)

git clone https://github.com/TensorLink-AI/cascade.git ~/cascade-upstream
cd ~/cascade-upstream
python3.12 -m pip install -e '.[hippius,chain]'

2. Install this repo

cd /path/to/cascade-subnet
pip install -e '.[dev]'

3. Observe testnet (no wallet)

cascade heat --network test --chain-toml ~/cascade-upstream/chain.testnet.toml
cascade-observer pull-round --config config/observer.example.yaml

4. Pack and verify

cascade-pack-submit --src generators/mixture --dest dist/submit
cascade verify ./dist/submit --chain-toml ~/cascade-upstream/chain.testnet.toml

5. Deploy (wallet + Hippius required)

See docs/TESTNET.md. Deploy dist/submit/, not generators/mixture/.

Repository layout

docs/           AI-oriented context (start at docs/INDEX.md)
generators/
  mixture/      Research mixture (not the Hippius push)
  pack_submit.py
observer/       Field log + fetch-field
labbook/        Experiment pages
dist/submit/    Packed artifact (gitignored)
config/         Example YAML
logs/           Local experiment logs (gitignored)
notebooks/      Operational Jupyter notebook

Documentation

Doc	Audience
docs/INDEX.md	AI agents — read first
docs/TESTNET.md	Operators — testnet checklist
docs/OPERATIONS.md	Wallets, costs, environments
docs/STRATEGY.md	Mixture / orthogonality strategy
docs/BACKLOG.md	Later rules and skills
labbook/	Experiment narrative

AI agents: read AGENTS.md.

anacoluthe89/test-auth-check

README.md