@OP: I agree with what everyone has said on these things:
- modern hoseless AI seems to be pretty solid. My Oceanic Atom 3.0 w/AI has been totally reliable for the 21 months and 100+ dives I've done with it. And I haven't even needed to change the transmitter battery yet.
- if you can wait until DEMA, you will probably find there are new options available that you might like better than what is currently available.
I would add:
Since you have so much experience and mentioned doing deep and deco dives, I personally would stay away from all the recreational computers, and go for something that will work well for deco dives.
I would not buy the Eon Steel because, while it does support Trimix, it does NOT use a tech-friendly algorithm.
The only computer that is available right now (that I know of), that has AI, that I would be looking at (and actually I am looking at as a backup to my H3) is the Hollis TX1. It's a wristwatch style, but bigger (more puck-sized). It is a full-on, multi-gas, trimix-capable tech computer that uses the Buhlmann algorithm, with Gradient Factors. And it works with the Hollis/Oceanic/Aeris AI transmitters. The computer is selling for something like $550. They want $950 for a package with the AI transmitter, but you can get the transmitter separately for well under $400. Or you could probably get the package for cheaper if you call CaveAdventurers or email LeisurePro.
As a photographer, I would think that being able to check your tank pressure by glancing at your wrist, versus having to take a hand off your camera (or whatever), or even turn your head away from your work, would be a useful advantage.
My backup computer/dive trip watch is a Suunto Stinger. I'm not quite sure why I carry it, though, because if my Cobra craps out, I won't have a tank pressure reading anymore...
If my Atom crapped out on a single tank dive, I would only have my backup computer and no tank pressure reading. I'd still be glad to have my backup computer. For one, it would still be telling me my depth and ascent rate. And, for two, if I'm using a single 120 on a wreck dive off NC, and I have a solid buddy, I probably wouldn't even end my dive early. I know my SAC well enough to be confident that my NDL will run out way before I get low on air. I usually max my NDL, take a long safety stop on the way up, and still get out with over 1000psi left (when I'm actually back on the boat).
Plus, once I'm back on the boat, if I had to, I could swap my AI transmitter off for a physical SPG and do the next dive with my backup computer having the correct info regarding my current tissue loading.
The decompression algorithm is simply more liberal.
Small correction to this. The IMPLEMENTATION of the algorithm is more liberal. An algorithm is an algorithm. No algorithm is inherently conservative or liberal, I don't think. It's the implementation of the algorithm that makes it liberal or conservative. For a bubble model, one implementation could use a different critical radius than another. For a Buhlmann model, one implementation could use the equivalent of gradient factors with different values.
And, that is what RonR has alluded to before, here (on page 4 of the thread, if the link doesn't go straight there):
New dive computer for Divemaster internship
I feel like it's important to keep the distinction clear between an algorithm and its implementation. Many discussions here blame an algorithm for things that should be blamed on the implementation. Which is to say that, for example, one RGBM-based computer being very conservative should not be an automatic whitewash of all RGBM-based computers as being conservative. You've pointed out the Cobalt as the prime example. Similarly, not all Buhlmann-based computers should be presumed to be liberal.
However, I *think* that DSAT is a specific implementation of a Buhlmann algorithm. So, I do think it is appropriate to expect all DSAT computers to be very similar in how liberal/conservative they are. However, I could be wrong. Any OEM that puts DSAT in their computer COULD (I conjecture) be free to tweak parameters to make their computer more or less liberal/conservative than other computers that are also using DSAT.