Night Shift Tokenomics

The research engine for systematic tokenomics analysis and resilience scoring.

Night Shift Tokenomics powers the Agentic Tokenomics Observatory by running large-scale, multi-gate simulations of Solana token economic designs.

What it does

• Explores tokenomic parameter spaces (fee routing, supply mechanics, treasury control, vesting, governance, etc.)
• Applies multi-stage validation: walk-forward analysis, overfitting gates, cross-regime consistency, Monte Carlo stress, and agent simulation
• Produces a Resilience Score and taxonomy of resilient vs fragile designs
• Exports Observatory datasets (manifest, dataset JSONL, rejections, regime counts, per-token detail)

Repository layout

Path	Purpose
SPEC.md	Full technical specification
src/night_shift/	Pipeline, validation gates, ingestion, batch runner, export
configs/	night_config.json (single runs), classification_batch.json (51-token batch)
scripts/	run_classification_batch.py, fetch_token_data.py
data/seed_tokens.json	51-token seed manifest (tracked)
tests/	Unit and integration tests

Runtime artifacts (data/exports/, data/tokens/cache/, data/tokens/backfill_manifest.json, data/batch_logs/) are gitignored and stay local.

Quick start

python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
pytest

Live backfill (Helius)

Set HELIUS_API_KEY in your environment (never commit it). Without a key, the pipeline falls back to label-conditioned bootstrap histories.

export HELIUS_API_KEY=your_key_here

# Phase A: fetch 270-day on-chain histories for all seed tokens
python scripts/run_classification_batch.py --phase backfill --config configs/classification_batch.json

# Phase B: classify and write Observatory export (v2_<date>/)
python scripts/run_classification_batch.py --phase classify --config configs/classification_batch.json

Use --force to refetch cached tokens. Use --run-id 2026-06-07 to pin the export directory name.

Single-token / dev runs

night-shift --config configs/night_config.json --tokens JUP

Status

Phase 2 in progress. Implemented and tested:

• Solana data ingestion (Helius REST, parquet cache, bootstrap fallback)
• Walk-forward grid search with coarse sampling
• Overfitting, cross-regime, robustness, and agent-stress gates
• 51-token batch backfill + classification + Observatory export schema

First public Observatory dataset is pending a full Helius backfill on real on-chain histories. Bootstrap-heavy runs are useful as internal baselines only.

Related projects

• Resilient Token Protocol (RTP) — https://github.com/tradewife/resilient-token-protocol
• Night Shift Security (parallel track) — https://github.com/tradewife/night-shift-security
• Website: https://www.resilientprotocol.xyz

Contact

Kate / tradewife
X: @trade_wife
GitHub: tradewife