A steady stream of brilliant, serious people are on stage warning that superintelligence is near, that we have a handful of years before everything changes, that this may be the last invention that matters. It is gripping, it is often thoughtful, and as a basis for running a business it is almost entirely useless, with one exception worth keeping.

Why most of it does not help you decide anything

Predictions about a transformed world in three years share a property that makes them useless as inputs: they are unfalsifiable on any timeline you can plan around, and they point in every direction at once. If the future is either utopia or catastrophe and nobody can say which, that is not a signal you can act on, it is a mood. Acting on the mood tends to produce two equally bad responses: freezing, on the theory that nothing you build matters, or lurching, on the theory that you must bet everything on the most extreme scenario. Both are worse than just doing good work. It is the same reason we argue for ignoring the consciousness question when the decision in front of you is operational.

The tell to watch for

The doom conversation runs on the same fuel as the hype conversation: maximum drama, minimum falsifiability, delivered by someone with something to sell, whether that is a book, a fund, or a worldview. Treat a stage prophecy the way you treat a breathless benchmark claim, as a starting point for your own thinking rather than a fact to act on. Most of it fails the only test that matters for an operator: does it change what your team can usefully do this quarter. That filter is the same one that cuts through ordinary AI news.

The part worth keeping

Here is the exception. Underneath the theatrics, the serious version of the argument is a case for building carefully: keep humans in the loop on consequential decisions, insist on systems whose actions you can see and undo, and do not hand irreversible power to something you cannot inspect. That is not doom, that is just good engineering, and we would hold to it whether or not the sky is falling. It is exactly why we treat accountability and inspectability as non-negotiable in our solutions.

So take the discipline and leave the prophecy. Build as if the near future is roughly like the present but with cheaper, more capable tools, because that is the version you can actually plan for, and keep the safeguards that would matter in any future at all.

Prompted by videos including “AI Superintelligence is coming” and long-form interviews on AI risk (2026). Credit to the speakers; the operator’s read is ours.