When prompt engineering makes you solve the issue before ChatGPT even answers

Description

Dark-mode screenshot of two stacked social-media posts. 1) Header row: circular avatar, name "Matt Novak", handle "@paleofuture.bsky.social", and a right-aligned "Follow" button. Matt’s text says: "AI folks have now discovered “thinking”". 2) Quoted tweet style box from Steph Smith: avatar, blue verification check, name "Steph Smith", handle "@stephsmithio", and its own "Follow" button. Her post reads: "Sometimes in the process of writing a good enough prompt for ChatGPT, I end up solving my own problem, without even needing to submit it." Underneath is a gray metadata line: "2:16 PM · 7/29/24 · 1.7K Views", plus a small rounded "ALT" badge. The outer interface footer shows "Jul 29, 2024 at 4:43 PM". Technically the meme riffs on prompt-engineering culture: articulating a precise question mirrors rubber-duck debugging, so developers often discover the answer before sending it to an AI assistant. It satirizes dependence on ChatGPT while celebrating the timeless benefit of clear thinking

Common Tags

Chatgpt AIAssistants AI DeveloperHumor DeveloperIrony

Comments

Anonymous ★ Top Pick Prompt engineering is just rubber-duck debugging billed per 1K tokens - plus the bonus risk that the duck hallucinates a new microservice

Anonymous ★ Top Pick

Prompt engineering is just rubber-duck debugging billed per 1K tokens - plus the bonus risk that the duck hallucinates a new microservice
Anonymous

After 20 years of explaining code to rubber ducks, we've finally achieved peak engineering: spending $20/month to almost explain our problems to an AI before realizing we already knew the answer. Next up: AI-powered cardboard cutouts of Linus Torvalds for premium debugging sessions
Anonymous

The real breakthrough in AI isn't the model's intelligence - it's that we've finally productized the rubber duck debugging methodology at scale. Turns out the most valuable feature of ChatGPT is forcing engineers to articulate their problem clearly enough that they solve it themselves before hitting 'Send.' We've essentially built a $10B company around the Socratic method with extra steps and a JSON API
Anonymous

Prompt engineering: rubber duck debugging, but the duck now charges per token and hallucinates less
Anonymous

LLM prompting is just rubber-duck debugging with embeddings - my RLHF now stands for Realizing Late I Had the Fix
Anonymous

Prompt engineering is just rubber duck debugging with a token meter - LLMs didn’t invent thinking, they just put a price tag on writing a decent spec
@paul_thunder 2y

Based
@RustyOtter 2y

https://en.wikipedia.org/wiki/Rubber_duck_debugging aka writing up the issue in an email and you either solve it along the way or you have an email to send to someone to ask for help
1. @Agent1378 2y
  
  I just talk loudly to myself😁
  
  @hotsadboi 2y
  
  i just whisper the code and add "but fucking why?" every line. never fails me
@azizhakberdiev 2y

Literally me whenever I try to write a question in SO
@DenDrobiazko 2y

ChatGPT is a perfect rubber duck. If you don't succeed in resolving the issue while explaining it can actually help. Also - look up perplexity.ai
@AlexAparnev 2y

Porn used to be simulation of sex. Now sex is simulation of porn. LLMs used to be simulation of thinking and talking. Now thinking and talking is becoming simulation of LLMs?