That’s a neat thought, but full autonomy means you’ve got to feed the AI all the data streams—weather, orbital mechanics, instrument health, even ground‑based calibration. You’ll need a learning loop that updates the pointing model on the fly, plus a fail‑safe that can bail out if the AI gets stuck in a local optimum or the sun comes in the wrong angle. If you can wrap that in a robust scheduler and an emergency manual override, you’ll get a sky‑roaming hunter that’s as reliable as a Swiss watch—minus the tiny watch‑band that gets tangled in the antenna cables.