Research · testnet only
Yahya AI Trading Lab
A research laboratory for trading hypotheses, with hard limits built into the code. It is an experiment, not a product.
- Area
- Experimental research
- Language
- Python 3.12
- Environment
- Spot testnet · paper only
- Status
- Research
Overview
YATL is where I test trading ideas before they are trusted with anything. It reads public market data and, with credentials, reads a testnet account. It cannot place orders.
My role
Sole author of the research code, the safety boundaries and the test suite.
Stack
- Python 3.12 using only the standard library
- uv for locked environments
- unittest for tests
- Binance Spot Testnet, read-only
Challenge
Research on trading systems tends to drift toward real money and toward unreviewed automation. The lab has to make the safe path the only path.
Approach
Boundaries are enforced in code, not only documented. Authenticated transport is limited to one GET endpoint. Configuration must say testnet and keep the master lock off, otherwise the program refuses to start. Redirects and ambient proxies are disabled, responses are size-limited, and credentials are redacted from every output.
The written rules are: paper only, no futures, no leverage, no withdrawal API, and no AI direct execution.
Gallery
Not yet documented. Screens or diagrams that can be shown publicly.
Verified outcomes
- The first test phase recorded 22 of 22 tests passing on 7 September 2026, including an RFC 4231 HMAC vector, credential redaction and redirect refusal.
- No trading edge has been demonstrated, and none is claimed.
Status
Research. The boundaries above apply to all of it.