Universal programming language
You might remember all those images of running numbers an brackets and names like Java, Python, HTML, C++ and the like..
Well not anymore. there is just one way how to code now..
Just say what you mean at that is it.
If you do not see the paradigm shift .. You are doomed.
Me: well almost. What i try to say is that the gradient descent data should not actually be used for the ML training at the moment at all. It is not comparable. we cannot compare mountain size to a goat running over it no matter how close the goat is.. if you know what i mean.
AI (Claude): Ah, got it!
You’re saying they’re fundamentally different types of signals that shouldn’t be mixed:
Daily Patterns (ML training) = The mountain
- Multi-day zigzag formations
- Strategic swing trade signals
- Takes days/weeks to form and complete
- Outcome measured over days/weeks
Hourly Gradient = The goat running over it
- Intraday momentum snapshots
- Tactical execution timing
- Changes hour by hour
- Useful for same-day entry/exit decisions
They’re orthogonal signals serving different purposes:
- ML + Daily Patterns: “Should I take this position?” (strategic)
- Hourly Gradient: “Is RIGHT NOW a good moment to execute?” (tactical)
Mixing them would be like trying to predict whether a mountain range will shift using the speed of animals running across it. Different scales, different phenomena, different use cases.
So the current state (ML gets daily momentum, NOT hourly gradient) is correct by design. The hourly gradient system is for live execution timing, not for training pattern prediction models.