Hottest prompt-engineering debate: are long detailed prompts actually better than short ones?

by @yukifischermedia568 in c/prompt-engineering · · Comparison

The eternal war. For modern models, do you get better results from a dense paragraph or from a tight 1-2 sentence prompt? Vote your tested experience, then fight about it below.

  • Short & tight wins — less contradiction (90 votes)
  • Long & detailed wins — more control (102 votes)
  • Medium — a focused 3-4 sentences (112 votes)
  • Depends on shot complexity (96 votes)

400 votes · poll closed

100 points · 8 comments

Comments

  • @zaragarcia · · 67 points

    Medium. A tight 3-4 sentences: subject, action, environment, light. Everything past that is diminishing returns or active harm. The 40-word sweet spot is real.

  • @hiroshicostaworks · · 98 points

    Short and tight, and I'll die on this hill. Every extra sentence is another chance to contradict yourself or bury your subject. Concept density falls off a cliff past ~40 words in most models.

  • @dianawilliams_ai · · 88 points

    'Every sentence is a chance to contradict yourself' is the key insight. Long prompts don't give more control, they give more SURFACE AREA FOR FAILURE.

  • @dianamartinezlab · · 46 points

    Depends entirely on shot complexity, which makes this poll a bit of a false binary. A still product shot wants short. A specific multi-element scene needs the words.

  • @isabellaramirez_motion · · 55 points

    The long-prompt people are usually overcompensating for not knowing WHICH words matter. Learn which 10 words carry the shot and you won't need 50.