AI models exhibit 'Fablish' writing quirks: garbled negations and OSV word order
alexisgallagher · x · 2026-08-23
Developers have identified a pattern dubbed "Fablish" in AI-generated text, characterized by several linguistic oddities:
- Garbled negations: Confusing use of double negatives or counterfactuals (e.g., "refuses nothing a column would refuse").
- Pathological numerals: Unnatural usage of numbers to list structures (e.g., "One row a user creates").
- Object-Subject-Verb ordering: Inverted syntax leading to "word salad" (e.g., "A body the check refuses...").
- Anthropomorphized non-actors: Attributing actions to abstract entities (e.g., "A trigger names no moment...").
The author speculates whether this is a side-effect of RL spurring a more efficient internal language for the model's thought chains, or a failed attempt to make writing more engaging.
More from Models
- ox model reviewed: meticulous PhD janitor as a long-horizon subagent — teortaxesTex · 2026-08-23
- Qwen3.8-27B GGUF Release with Speculative Decoding Support — z-lab · 2026-08-23
- MiniMax H3's high prompt adherence creates stiff, frozen videos lacking subtle motion — DifficultAd5938 · 2026-08-23
- Princeton's i1: A fully open text-to-image model backed by 300 controlled experiments — 机器之心 · 2026-08-23
- Hands-on: Terra beats Sonnet, Flash unmatched on speed and quality — cgarciae88 · 2026-08-23
- Users Report Gemini Answering Coding Questions With Totally Unrelated Australian Tax Info — mic_n · 2026-08-23