Repeating instructions in prompts still makes models adhere better, like it's 2023
tokenbender · x · 2026-09-08
The author observes a counterintuitive throwback: people are discovering that repeating instructions within prompts makes models adhere to them better — 'like it is 2023.' A reminder that instruction-following in long contexts remains imperfect and old prompt-engineering tricks still work.
More from Models
- Claude Code 20x Max at $400/month equals ~$16,000 of API usage — tedddyoweh · 2026-09-08
- AA index updated: Fable 5.1 and GPT-astra now tied — jietang · 2026-09-08
- Video editor says GPT-6 Astra rebuilt his full Premiere Pro edit in minutes from raw footage — _AustinCalvert_ · 2026-09-08
- Gemini 4.0 teased as Google reportedly prioritizes Flash models over a Pro flagship — haider1 · 2026-09-08
- Don't hand off to a cheap model: Luna fails as a subagent for Astra on tasks needing real understanding — brandon_galang · 2026-09-08
- Real-world test punctures 'AGI has arrived' hype: Astra botches a simple website task — GaryMarcus · 2026-09-08