Yes, a strategy does exist which can force at least one line in a well of width 10 (and different strategies for 4, 6 and 8). I did a tedious brute-force calculation which demonstrated this a few years back. The code is not particularly presentable and the calculation takes about 11 days on a modern machine (it runs in a single thread), but here: https://github.com/qntm/tetris
The limiting factor is headroom (the depth of the well). For wells of depth 0, 1 and 2, the AI clearly wins, but once the depth increases enough a strategy becomes possible. My belief is that such a strategy exists for all even well widths, although proving this is annoyingly hard.
I find one of the most fascinating questions about constructing an escape room is how exactly you calibrate the difficulty for different groups of people. Some people will finish the room in 30 minutes like you, but others will be totally stumped after 50 minutes despite it being the exact same room.
Both of the rooms I've done had a hints system so a staff member overseeing the game through cameras could send a hint any time we dropped significantly behind schedule.
One of them had a two-room system where you had to solve the first room to get to the second room, and in the second room I believe one or two of the locks had been unlocked ahead of time manually by a staff member, based on how quickly we solved the first room.
Of course every time anybody takes the challenge you acquire more data, and keeping an eye on the emerging patterns must be absolutely critical to engineer an enjoyable challenge for the maximum number of different people. Metrics!
If you're puzzled about the implementation, I invite you to inspect the JavaScript code! And if you disagree with my implementation, I'd be interested to know what you would do differently.
This is a question of implementation, and answering it was part of the reason why I implemented this problem in the form of a computer program in the first place. In this implementation, the True and False gods simply pose the question to Random and take Random's answer - which is random. If you then ask the same question of Random again, of course you may not get the same answer.
> He proudly announces that there are ‘no fewer than 147 Indian dialects’ – a pathetically inaccurate count. (Today, India has 57 non-endangered and 172 endangered languages, each with multiple dialects – not even counting the many more that have died out in the century since My Fair Lady took place)
So, how many were there really? At the time, I mean.
It would be good to have a rigid specification for this kind of thing, same as for the colour and blink frequency of signal lights, road sign designs and number plate fonts. Predictability and standardisation is super valuable on the road.
However, I see no reason why the specification can't specify a particular electronic burbling noise from The Jetsons.
In the context of the story, information is a substance that you can almost shovel around like snow. Brain-uploading isn't an inevitability in the real world but in Fine Structure it's a (relatively) simple application of informational plumbing.
The limiting factor is headroom (the depth of the well). For wells of depth 0, 1 and 2, the AI clearly wins, but once the depth increases enough a strategy becomes possible. My belief is that such a strategy exists for all even well widths, although proving this is annoyingly hard.