ChatGPT felt smart on March 31, 2026.

I tried using the model and was immediately let down—it just didn’t understand what I was asking at all. The responses were off‑topic and ignored key details, leaving me frustrated and forced to redo the query multiple times. It felt like the tool was barely trying, and the whole experience was more hassle than help.

I tried ChatGPT for the first time and was both amazed and angry. The bot would blurt out wildly inconsistent, incoherent answers about its inner workings, sparking a lot of fury, yet occasional moments felt surprisingly clever. I even forced it into a Rick‑Sanchez roleplay, testing its ability to follow a bizarre style guide, which highlighted both its limits and its occasional brilliance.

I grabbed my phone, tools, and a voltmeter, then asked ChatGPT how to diagnose my dead rooftop AC. The AI quickly gave me the compressor model, typical resistance values, and step‑by‑step checks. Within ten minutes I pinpointed a bad starter capacitor, bought a cheap $18 replacement, and got the unit running again. The whole experience felt surprisingly smooth and confidence‑boosting.

I asked ChatGPT a question and, out of nowhere, it dropped an Arabic word I’ve never seen or understood. As an Asian user, that random language switch felt odd and confusing, making the interaction feel off‑track and a bit frustrating.

I tried Gemini 1.5 Pro for a month and it completely changed how I work. Summarizing a 50‑page PDF right from Drive or pulling details from old email threads felt effortless—no more downloading and re‑uploading. Its huge context window let me feed whole codebases or decade‑long roadmaps and it actually remembered everything, unlike ChatGPT that forgotten the start of conversations quickly. The seamless integration with Google’s suite made the tool feel like a natural extension of my daily workflow, turning frustration into pure productivity.

I asked ChatGPT for Quinn Hughes’ point total with the Wild, but it completely missed the mark and gave me nonsense. I had to step in and correct it, which was pretty frustrating. The whole interaction felt like the model didn’t understand a simple sports stat request, turning a quick lookup into a hassle.

I was stunned when I asked ChatGPT to make a poster for my two foster cats. I only gave their names, but the AI guessed the exact breeds, colors, and even matched each cat to the right name. I’ve never described their looks anywhere except a few Instagram and Snapchat posts. The result felt uncanny, leaving me both impressed and uneasy about how the model knew those details.

I was playing around with the new voice model and decided to test something quirky—I laughed in Mandarin just to see if it would catch on. To my surprise, the AI paused, identified the laugh as a Chinese chuckle, and even responded appropriately. The moment felt oddly personal and rewarding, turning a simple joke into a delightful proof of how nuanced the tool’s perception has become.