🤖 AI Scanner

The other five scanners are deterministic — fixed formulas that fire when their exact conditions are met. The AI Scanner is the flexible one: instead of parameters, you give it instructions in plain English ('only trade pullbacks in uptrends, skip anything before major news, prefer 1:3 or better'), and Claude — the AI model by Anthropic — reads recent price action against your rules and decides.

Every proposal comes with an entry, stop-loss, target, and a written rationale, so you can see why it wants the trade — and it either alerts you or trades automatically inside the same risk limits every Poshkan scanner obeys.

How the strategy works

1. Write your strategy in plain English — or start with the built-in one and edit from there.
2. On each run, Claude analyses your chosen symbols and their recent price action against your instructions.
3. It proposes trades with an entry, stop, target, and a written rationale.
4. In alert mode it notifies you and does nothing else; in auto-trade mode it opens positions within your risk %, max-open, max-per-day, and daily-loss limits.

When it shines — and when it doesn't

It shines when your edge is judgement rather than a formula — nuanced rules, context, exceptions — or when you want to prototype a strategy idea in an afternoon by literally writing it down. It's also the honest one about its limits: because the model isn't deterministic, it can't be backtested the way the formula scanners can.

Key terms

Deterministic vs discretionary

Fixed-formula strategies always fire identically on the same data; discretionary ones apply judgement — the AI Scanner is judgement, encoded in your words.

Rationale

The written 'why' attached to each AI proposal — the part a signal service never gives you.

Risk limits

Hard caps (risk %, max open trades, daily loss limit) that bound the AI regardless of what it wants to do.

How to judge it

Since backtesting doesn't apply, judge it live: let it run in alert mode on virtual money, read its rationales, and track how its proposals play out before ever enabling auto-trade. That workflow — try, watch, then trust — is exactly what a paper-trading platform is for.