tl;dr:
- Prompt engineering is very much alive.
- “Few-shot prompting” can improve accuracy from 0% to 90%.
- Role prompting (e.g. “You are a math professor. . .”) is largely ineffective.
- Advanced techniques like decomposition and self-criticism unlock better performance.
- Context (“additional information”) is underrated—and massively impactful.
- Real leverage comes from crafting high-performing prompts inside products. These prompts are used at scale, run millions of times, and must be hardened and optimized like production code.
joe
npub1dq…f2wpa
2025-06-19 18:21:09 UTC
Author Public Key
npub1dqn772m4aejjmnyrjk9c8t4qh3jcquzm2czp48h8pfqh3cgydnds8f2wpaSeen on
wss://nostr.oxtr.devPublished at
2025-06-19 18:21:09 UTCEvent JSON
{
"id": "c2e1af6c151c4d296ed2be1d59940246515c96f2c1cb5a8fd0883518e1b8d2c2",
"pubkey": "6827ef2b75ee652dcc83958b83aea0bc6580705b56041a9ee70a4178e1046cdb",
"created_at": 1750357269,
"kind": 1,
"tags": [],
"content": "tl;dr:\n- Prompt engineering is very much alive.\n- “Few-shot prompting” can improve accuracy from 0% to 90%.\n- Role prompting (e.g. “You are a math professor. . .”) is largely ineffective.\n- Advanced techniques like decomposition and self-criticism unlock better performance.\n- Context (“additional information”) is underrated—and massively impactful.\n- Real leverage comes from crafting high-performing prompts inside products. These prompts are used at scale, run millions of times, and must be hardened and optimized like production code.",
"sig": "c298d0afb3ef350ac4b6cde5a2203f696110feca08327db839defe863711e96b6248a4dd2bc51da3b48ecf1920a1253610f4f287d02b7bfbd57e41a477173a56"
}