Why Amaneki

The honest case for subscribing to a regime API rather than rolling your own or buying from an incumbent.

vs. DIY (you write it yourself)

A 3-state volatility FSM is a couple hundred lines of Python. Anyone with Binance API access can ship one in a day. What you're paying Amaneki for is the next nine months: running the poller 24/7, absorbing exchange downtime, keeping thresholds calibrated as the market shifts, reconciling Binance / Coinbase / Bybit symbol quirks, handling the WebSocket fan-out. The signal is the easy part; the uptime is the hard part.

vs. Glassnode / Coinglass / Kaiko

They're broader — hundreds of metrics across on-chain, derivatives, flow. We're one metric done deeply: regime detection with per-symbol calibration, cross-exchange consensus, and an RV-vs-IV gap for BTC/ETH. If you need the full quant warehouse, use them. If regime is specifically what you're optimising for, we go further on that single axis.

vs. AI-generated code

A language model will write the detector for you in 30 seconds. It won't write the backfill, the Postgres migrations, the signing scheme on your outbound webhooks, the rate limiter, the CI pipeline, the status page, or the on-call when Binance rate-limits you at 3am. What AI compresses is the code-writing. What we compress is the operational commitment.

vs. TradingView alerts

TradingView is for humans watching charts. Amaneki is for programs reading JSON. Different tool, different consumer.

Honest limits

We publish F1 0.39–0.45 at the balanced preset, per symbol. That is a noisy-problem number, not an impressive one. Regime detection against a smoothed percentile ground truth has a ceiling around there; anyone advertising F1 > 0.9 is labelling with future information.

We don't forecast price. We don't tell you what to trade. We tell you what state the recent vol is in, as cleanly as we can measure it.