Pieza independiente

The hypothesis queue

Movement 5 · The Laboratory That Watches Itself — September 2026

I What it does and what it does not

In February 2025 a system appeared with a clear function: it generates hypotheses from the literature, sets them against each other in something like a tournament, has other agents criticise them, and returns a ranked list.

It runs no experiments. It measures nothing. Its product is a priority queue [ordered, argued and very convincing].

That makes it an allocation instrument, not a discovery one.

“Deciding what gets investigated is a way of deciding what will be known.”

II Where each thing lands

Kind of hypothesis	Usual position	Reason
Consistent with the corpus consensus	top	the literature backs it
Cheap to evaluate	top	the scorer rates it well
Against the consensus	bottom	it contradicts the input material
Needing odd instrumentation	bottom	nobody can score it
Wrong mechanism, good data	very bottom	it loses on the explanation

That last row has a name in history. Semmelweis had the table of deaths by ward and by month [two years of records] and he also had a false mechanism. A queue ranked by theoretical plausibility would have buried him.

III Rules of use [safe]

Look at the input corpus before you look at the output. A queue is as biased as its library.
Require the system to show why it ranked that way, not only how.
Reserve a fixed share of the budget for the bottom of the queue; without a quota, exploration disappears.
Write down what it proposed and what turned out to be true. Without that record there is no way to know whether it works.
Keep the discarded hypotheses with their dates. The discard is material too.

IV Risks

The big risk is not the one people fear, that it proposes nonsense. It is that it proposes reasonable things all the time, and that a whole field spends a decade confirming what already looked true.

There is a second, quieter risk. The system trains on published literature, and published literature is skewed towards what worked. Negative results barely exist in the corpus; the queue inherits the structural optimism of a profession that does not publish its failures [and promotes the people who are right].