Prompt Fix for AI's "Machine Dialect": Split Long Sentences, Keep Connectives
xiaohu · x · 2026-10-27
xiaohu turns the NYU/UT Austin linguistics findings into a practical fix: to make LLMs write readable prose, you must actively counter their "machine dialect" in the prompt.
Concrete instructions to add:
- Break up long, complex sentences
- Restore subordinating words (that/which)
- Keep necessary prepositions and transitions
- Ban rare, invented noun compounds
The goal is re-translating high-density text optimized for code and reasoning into language that matches human cognitive rhythm. Full write-up and paper links included.
More from Apps
- Querio launches: analytics platform that puts your data stack in a git repo for AI agents — thetripathi58 · 2026-10-28
- AI Mode builds an playable synth on the fly to teach how synthesizers work — rmstein · 2026-10-28
- Parallel launches Parallel Decisions API at $0.001 per request with 5,000 free calls — schwentker · 2026-10-28
- Grok Bot demo shows new Pages, Vault, 1Password, Apple integrations and video generation — EricBuess · 2026-10-28
- Obsidian search: pick your tool by what you remember — a practical decision thread — dSebastien · 2026-10-28
- Solo dev runs a fully AI-generated YouTube channel with HeyGen and Claude Code at ~$20/video — pavlenkovit · 2026-10-27