Kiritasu: Tell AI exactly what to see | Orynth