Medium LLM
7/10 signal
Claude Fable 5 and the Inversion of Prompt Engineering: Why Your Best Prompts Now Make It Worse
reasoningcontext
What happened
The article discusses a shift in prompt engineering for advanced reasoning models, highlighting how over-specifying instructions or adding manual chain-of-thought steps can actually degrade performance compared to simpler prompts.
Why it matters
As models get better at internal reasoning, legacy prompt engineering techniques can actually degrade performance.
The take
This aligns with the broader trend seen in models like OpenAI's o1/o3 and Claude 3.7. Legacy prompt engineering techniques (like manual CoT) often clash with the model's internal reasoning loops, requiring developers to unlearn old habits.
Do this
Simplify your prompts and remove manual chain-of-thought or step-by-step instructions when upgrading to advanced reasoning models.
Don't read this site daily. Get it in your inbox.
The daily brief and Sunday deep dive — distilled, scored, and opinionated. For builders only.