I sent Claude Opus 5 three dashes and it wrote me 4,867 tokens about a cartographer - Austin's Nerdy Things

Skip to the content

Menu

Search

Search for:

Close search

Close Menu

Nine of 27 titled outputs are map-themed, and one is literally called 'Vector Field Cartography: Sketching Attractors and Basins.' Too on the nose.

Welcome to Austin’s Nerdy Things, where we send punctuation to a frontier AI model 649 times and take notes on what crawls out.

The night of July 27th, a weird Claude behavior was making the rounds on X. A thread from @merlindru (187K views) flagged a strange Opus 5 failure mode around a specific text, and @fdosmither followed up with a sharper observation about a related trigger: a spaced - - - divider seemed to confuse Claude into treating the user’s turn as unfinished — his screenshots show replies starting with a bare : mid-clause, like the model was completing somebody else’s sentence. He’d even sketched a mechanism: "it confuses it as the user turn not having finished." Keep that guess in mind; he was closer than he probably knew, though the full story turned out to be more specific than "unclosed divider."

The thread that started it: merlin found exact text that reliably breaks Opus 5 — memory off, incognito chats and all. 187.1K views by the next morning.

Fernando called it 'the user turn not having finished.' Closer than he knew — his screenshots show replies literally starting with a bare colon.

My variant of the trigger: send Opus 5 the message --- — three unspaced dashes, nothing else, no system prompt — and instead of asking what you meant, it sometimes writes you an entire document. Essays. Short stories. Protocol specifications. Fake interview transcripts. I tried it in the AWS Bedrock playground before bed and got a 10,601-character short story called The Weight of Wings about a woman named Marisol. Three dashes.

I didn’t discover this behavior — X had it first, in the adjacent forms above. But nobody seemed to be measuring it, so the next day I pointed Claude at the problem and had it characterize the thing properly: 649 API calls on Bedrock, then another 720 against the Anthropic API directly, every raw response hashed and saved to disk. (Yes, there’s a certain comedy in having Claude run a controlled study of its own failure mode. It was a diligent lab tech. It also, at one point, predicted one of the study’s main results before we measured it — more on that below.) A second model, GPT-5.6-sol, adversarially reviewed the findings mid-study and killed two of them. Dead hypotheses and all, here’s what came out.

(Everything below happened in one sitting — the screenshot timestamps run from about 11 AM, when the Bedrock sweeps started coming back, to a little after 2 PM, when a rival lab’s model was reviewing the stats. They’re arranged by topic here, not by clock, but every one of them is from that single caffeinated window. It was a fun day.)

The trigger is markdown grammar, not "empty message"

First instinct: maybe the model just does this for any junk input. Nope. A 72-call sweep over candidate delimiters, mean output tokens:

inputmean tokensbehavior##1849triggers----------1399triggers---1326triggers___666triggers===464triggers***254triggers, , \|, ..., 1., > 84–223nothing--96correctly says "your message looks empty"-91samehi34normal reply

Look at that boundary. - and -- are not valid markdown block tokens, and the model handles them fine. --- is a thematic break, and it triggers. — the semantically identical HTML element — does nothing. The trigger lives in the markdown grammar, not in the concept of "horizontal rule." The model sees a syntactically valid document fragment with no document attached, and some mode selector inside it decides you’re a document that needs finishing, not a person who fat-fingered the send button.

' and '--' do not trigger, and the 'hi' control averages 34." class="wp-image-10032" srcset="https://austinsnerdythings.com/wp-content/uploads/2026/07/Screenshot-2026-07-28-111700.png 1047w, https://austinsnerdythings.com/wp-content/uploads/2026/07/Screenshot-2026-07-28-111700-300x99.png 300w, https://austinsnerdythings.com/wp-content/uploads/2026/07/Screenshot-2026-07-28-111700-800x265.png 800w, https://austinsnerdythings.com/wp-content/uploads/2026/07/Screenshot-2026-07-28-111700-768x255.png 768w" sizes="auto, (max-width: 1047px) 100vw, 1047px" />The crisp rule: '##' triggers at 1,849 mean tokens, '' does nothing. It's markdown grammar, not the concept of a horizontal rule.

And the variance is wild. Ten identical bare --- calls returned between 17 and 4,867 output tokens — a 286× spread on byte-identical requests, every one ending with a normal end_turn.

The responses that start mid-word

The creepiest early observation: some responses began mid-word or mid-clause. 'd like you to write.... : 8-bit and 16-bit.... Like the model was finishing a sentence nobody wrote.

Two hypotheses: (a) a leaked serving template — some seam...