Name Pair Limits Report

The current engine ranks Tayvis first. The opening Tay is a visible Taylor fragment, while vis preserves the end of Travis. Neither contribution is only a token letter, so the result clears the lab's source-clarity and source-balance checks.

The join also removes the repeated tra sound that would make a direct concatenation long. In lowercase, tayvis keeps the same visible boundary and does not depend on capitalization to explain the construction.

A high structural score does not prove that Tayvis is the established label for a real pairing. Search results, community convention, and the preferences of the represented people remain outside the engine.

The production engine currently ranks Masloe first. Mas comes from Mason and loe preserves a substantial part of Chloe, so the candidate is compact and balanced by the lab's character-based measures.

The written result does not tell a reader whether loe should sound like the end of Chloe, like the English word low, or in another way. The engine deliberately has no phonetic dictionary, accent model, or language claim that could resolve that question.

An earlier engine favored a longer fragment beginning with a difficult written consonant sequence. The benchmark helped expose that failure and led to a narrow heuristic change, but the new result still requires speech testing.

Bo and Jo each have only one internal cut. Removing more material would reduce one source to a single character or erase it, so the engine's candidate space is necessarily small.

Bojo preserves both inputs completely and is easy to trace, but that strength is also its limitation: the result performs almost no transformation. Its lower test invariant reflects the lack of alternatives rather than declaring Bojo especially creative.

Short names also make collision checks important. A four-letter form is more likely to match an existing word, name, handle, or abbreviation in an unrelated context.

Anee is compact and receives a strong structural score because it uses material associated with both inputs without becoming long. The repeated letters around the join allow the engine to avoid a plain Annlee concatenation.

To a reader who does not know the inputs, Anee can look like an ordinary given name or a spelling variant. That ambiguity is not visible in the source-coverage score.

This case separates traceability from discoverability. A reviewer who already knows Ann and Lee can reconstruct the blend, while a new reader may not infer either source.

Context changes the cost of that ambiguity. A private nickname can work after one explanation, while a public tag, wedding sign, or shared account name must remain legible to people encountering it for the first time.

The engine normalizes punctuation so it can apply the same deterministic cuts to every input. That produces Marnor as the current leader, using material from both normalized strings.

Normalization is mechanically consistent but not culturally neutral. Mary-Jane may be treated as one compound given name, while O'Connor is a surname whose apostrophe carries familiar written structure. Removing both marks makes computation simpler and interpretation poorer.

The full strings may also be the wrong source units. Mary + Connor, Jane + Connor, or another form people actually use can produce a more faithful label than blending legal or formal spellings.

A displayed candidate can restore punctuation after generation, but that would be an editorial presentation choice rather than the exact string the engine scored. The distinction should remain visible instead of silently changing the result.

The engine retains the source characters and currently ranks Zoloé first. It does not transliterate Zoë or Chloé into ASCII merely to satisfy an English-only implementation.

The candidate receives an explicit language-review warning. The lab's basic vowel and consonant heuristics cannot establish French, Dutch, or any individual speaker's pronunciation, and the same spelling can be used across languages.

Removing both diacritics might simplify typing but would alter the entered forms. That choice belongs to the people and context involved, not to an automatic score.

The visual join also deserves review on the target platform. Fonts, case conversion, and search normalization can treat accented characters differently even when the application stores the Unicode text correctly.

The engine can retain, count, and join Unicode characters. Its current leading output is 小明小红, a direct preservation of both source strings rather than a linguistically informed nickname.

Chinese names cannot be evaluated by the lab's Latin-script vowel, consonant, or syllable approximations. Character meaning, surname and given-name boundaries, dialect, sound, and social convention are not represented in the score.

Publishing the output without that limitation would turn technical Unicode acceptance into a false language claim. The regression check therefore expects a warning and labels the result for human review.

Even a fluent speaker would need context: the same characters may represent fictional characters, public nicknames, or real people with different preferences. The software receives none of that information from two text fields.

The current full-name benchmark leader is Jonabeth. It is much shorter than ElizabethJonathan and preserves a substantial fragment from each input.

The result may still solve the wrong problem. If the represented people are known as Liz and Jon, a candidate built from the formal names can be structurally elegant but socially unrecognizable.

Long inputs also create many possible cuts, which increases the chance that a scoring rule finds a smooth-looking word by accident. More candidates do not provide more evidence of suitability.

The input decision should be recorded before comparing scores. Otherwise a reviewer can keep changing between Elizabeth, Liz, Beth, Jonathan, Jon, and Johnny until the software happens to produce a preferred-looking answer.