O'Reilly: Coding With GPT

Content movement _directi rotate_fly(

Left guy's in denial about AI coding tools while right guy's already seen the git blame logs from the future. The eternal cycle continues: new tech emerges, developers panic, then end up maintaining the mess it creates. The only constant in programming is cleaning up after the latest "revolutionary" tool. Just wait until we're all writing prompts to fix the prompts that fixed the code that broke the system.

When someone fixes OpenAI's UI bug and asks for $5/hour plus a cola, you know the compensation expectations have been properly calibrated by years of API bills. The left side shows the broken error message getting cut off mid-sentence, while the right shows it actually readable. The irony here is delicious: this person is literally offering to work for less than minimum wage at a company that charges you $20/month for ChatGPT Plus and burns through your API credits faster than you can say "token limit exceeded." They've seen enough "Add credits" popups to know that OpenAI has no shortage of revenue. Honestly though? Fixing UI bugs that actually improve user experience is worth way more than whatever equity-heavy, ping-pong-table-adjacent offer they'd get. The cola is just chef's kiss.

Oh. My. GOD. The absolute TRAGEDY of modern development! 😭 You start with ONE measly problem that's driving you insane. Then your sleep-deprived brain thinks, "ChatGPT will save me!" But PLOT TWIST! Now you're drowning in a sea of strawberry-themed errors because ChatGPT decided to hallucinate an entire fruit salad of solutions! Is it "strawberry" or "strawbery" or "strawberrrrrry"?! WHO KNOWS ANYMORE?! The AI doesn't! And now instead of fixing one problem, you're questioning your entire career choice while debugging AI-generated code that's about as reliable as a chocolate teapot. The circle of developer suffering is COMPLETE!

Modern programming is just AI shopping. Why solve a problem yourself when you can make five different AIs race to solve it for you? The real skill isn't writing code anymore—it's knowing which AI's hallucinations are least likely to crash in production. The "hit run on all five" part is where the true chaos begins. Nothing says "I trust my code" like throwing five different AI solutions at the wall and seeing which one sticks. Bonus points if you don't actually understand any of them but confidently present the winner in your next code review. The "like a psychopath" is just chef's kiss perfect. Because what's more psychotic than spending 3 hours asking AIs the same question when you could have just written the damn function yourself in 20 minutes?

AI is that new junior dev who can write an entire e-commerce platform from scratch overnight but stares blankly at the screen for hours when you ask it to fix a missing semicolon. The duality is real - fancy tuxedo for creating complex systems, but turn into a teeth-grinding maniac when faced with debugging the simplest issues. Just like the rest of us, except we don't have the excuse of being a neural network.