A dive computer measures exactly two things: depth and time. Everything else on the screen is inference.

What the model is doing

Decompression algorithms divide you into notional tissue compartments, each absorbing and releasing inert gas at a different rate, and track the loading of each one. Fast compartments matter on a short deep dive; slow ones matter across a week of repetitive diving. Your no-decompression limit is the point at which the model says one of these compartments is close to a threshold — a threshold derived from population data, not from you.

Which is why two computers disagree

Put a Bühlmann-based computer next to a reduced-gradient bubble model on the same dive and they will give you different remaining times. Neither is broken. They encode different assumptions about how bubbles form and how conservative to be. This is a normal fact about models, and it is the reason the answer to "which computer is right" is that the question is wrong.

What it cannot see

• You. Hydration, sleep, fitness, fat distribution, a cold, a hangover, age. None of it is an input.
• Your workload. Fighting a current loads gas faster than drifting through the same profile.
• Yesterday, properly. Repetitive-dive handling is a simplification, and a week of aggressive profiles accumulates in ways the model only approximates.

So use it as an instrument

Which means: add conservatism deliberately rather than diving to the last minute of the display, ascend slower than it demands, and treat a no-decompression limit as a boundary you plan to stay well inside — not a target to spend. Every diver I know who has been bent was inside their computer's limits at the time.

I have the same relationship with a monitoring dashboard. It tells me something true about a model of the system. It does not tell me the system is fine.