Carter Church’s September 18 writeup reports an AI-assisted reading of a Napoleonic cipher. The interesting part is the inspectable evidence package, not simply a claim that an AI solved an old mystery. This article explains how a reader can separate transcription, decoding and historical interpretation, with unresolved signs kept visible rather than polished away.
Uncertainty remains visible
The writeup includes unresolved signs and alternative readings. That is important context for the word solved: a broadly coherent decoding need not make every mark certain. Our interpretation is that a reader should inspect the mapping and the remaining ambiguity separately from the narrative translation. A fluent modern rendering can conceal editorial choices. This article does not reproduce the letter’s full translation, make a new cryptographic claim or establish that an AI can break arbitrary encryption.
Three layers to examine
Our proposed worksheet starts with image-to-symbol transcription, then symbol-to-text mapping, then historical interpretation. For each disputed passage, identify which layer creates the uncertainty. A symbol that is difficult to read is not the same problem as a confidently read symbol with an uncertain value. Ask whether an interpretation remains plausible under the documented alternatives. Keep the literal decoded text apart from a polished translation, so an editorial improvement cannot accidentally be presented as extra evidence.
What a reproduction should report
If you inspect the package, retain its version and checksum and record which script or manual check you performed. Test the key against the recorded transcription before making claims about the scan itself. A successful scripted regeneration demonstrates consistency between supplied files; it does not independently authenticate their transcription. For a stronger review, compare selected signs against the original source image and document disagreements. We have not performed those checks, and readers should not interpret this explanation as a completed archival or cryptanalytic audit.
The broader lesson for AI research
This case is useful as an example of an answer accompanied by inspectable intermediate artifacts. When covering other AI claims, we would ask for the same separation between original input, processing decisions and final narrative. Avoid treating time spent by a model as a quality score, or replacing expert scrutiny with repeated chatbot agreement. The formal mathematics guide concerns a different domain but shares the principle: readers need the artifact and its limitations, not only the impressive outcome.
Frequently asked questions
Was the writeup published today?
No. Carter Church dates it September 18, 2026; this explainer is new coverage.
Did FyreLinkz independently decode the letter?
No. We reviewed the writeup and did not run or audit its solution package.
Are all signs certain?
The author displays unresolved signs and alternatives; preserve that qualification when describing the reading.
Sources & further reading
Primary sources checked Oct 11, 2026. Vendor statements are attributed; editorial advice is our own.
- 1Breaking the Marmont Cipher, 1809 ↗Carter Church
Help us keep this useful. Check contact availability →
