In 1931 a logician showed something narrow and startling about the kind of formal systems mathematicians use to prove things, and the result has since been dragged into every argument its precision was designed to prevent. It is worth stating carefully before it is used for anything else.
1. A sufficiently expressive, consistent formal system contains true statements about numbers that the system cannot prove.
This is the first incompleteness theorem, and it applies to formal systems of a particular kind: ones expressive enough to talk about basic arithmetic, built from fixed axioms and mechanical rules of inference, the sort of scaffolding mathematicians use to guarantee that a proof, once accepted, is beyond dispute. Inside such a system it is possible to construct a sentence that, in effect, says of itself "this sentence cannot be proved here." If the system is consistent, meaning it never proves both a statement and its negation, then that sentence is true and unprovable at once. Not false and unprovable. True. The system is not broken. It simply cannot reach everything that is so within its own reach.
2. Such a system, under standard conditions, cannot prove its own consistency.
This is the second theorem, and it follows from the method used to build the first. A system powerful enough to reason about arithmetic can, in principle, formulate the very statement "I am consistent." What it cannot do, if it really is consistent, is prove that statement using only its own rules. To vouch for its own reliability it would need a vantage point outside itself, a stronger system, and that stronger system would face the identical problem one level up. There is no floor.
It is tempting, reading this, to reach immediately for grand conclusions: that certainty is impossible, that no belief can ground itself, that reason has finally been shown its limits. Resist the reach. Goedel's theorems concern formal systems of a specific and technical kind, systems with fixed axioms and mechanical rules, built to talk about arithmetic. They say nothing directly about ethics, or love, or whether a novel is any good, or whether a religious tradition is true. The theorem is precise exactly because it stays inside its lane.
3. Something in the shape of this problem echoes, as analogy only, in older doubts about self-justifying belief.
Long before Goedel, thinkers wrestled with a related shape: can a mind trust its own faculties using only those faculties, or does trust always require standing on ground the mind did not itself lay down. A method of radical doubt tries to find one thing that cannot be doubted, and even that search has to use the very reasoning it is trying to certify. This is not the same problem as Goedel's, and it would be wrong to say the mathematics proves the philosophy. The kinship is structural, not logical: both point at how hard it is for a system, formal or mental, to step fully outside itself and check its own foundations from nowhere.
4. Incompleteness is not a failure of the system. It is a description of any system honest enough to be checkable at all.
The theorems apply only to systems strong enough to be interesting, systems that can actually do arithmetic and prove real things. A trivial system, too weak to say much, can sometimes prove its own consistency, precisely because it cannot say enough to trap itself. Power and self-containment trade off against each other. This is closer to a fact about the architecture of formal reasoning than a tragedy, though it has been narrated as tragedy more than once.
What stays with me is not the limit itself but where it sits: not at the edge of what a system knows, but at the edge of what it can know about itself while remaining inside itself. Every proof needs a place to stand that it did not itself prove.