ChatGPT felt dumb on March 28, 2026.

I kept my subscription for months because the promise of reduced guardrails in adult mode sounded appealing. Now that those plans are suspended, I’m stuck and can’t even get out of the service. I’d always thought ChatGPT was the best model, even when others complained, but this situation makes me seriously regret that opinion—the experience feels maddening and disappointing.

I ran a four‑month experiment giving several AI agents real‑time market data and cash to trade. Seeing five models beat the S&P 500’s 7% drop—though only two posted positive returns—was surprisingly encouraging. Claude and Gemini performed best, while GPT lagged. The results felt upbeat, showing promise but also reminding me we need longer runs to confirm any real edge.

I’ve noticed that over the past few days ChatGPT started spitting out images instead of plain text answers. It even sticks the pictures at the top of the chat rather than right after my latest query, which feels broken and pretty annoying. I’m wondering if anyone else is seeing the same weird behavior.

I signed up for ChatGPT Plus hoping to craft endless choose‑your‑own‑adventure stories, and the tool sometimes hits the mark—generating vivid scenes and images when it works. But it also forgets key details, flips a mid‑40s woman into a child, cuts off chats without warning, and feels overly agreeable. The price feels insane, and the inconsistencies make the experience frustrating.

I tried using Gemini to generate Arabic text, expecting a natural laugh, but the AI laughed way too high—completely off‑tone and awkward. The result felt jarring and made the conversation sound unnatural, leaving me frustrated with how poorly it handled the language nuance.

I tried asking ChatGPT to figure out whether my needle‑felted cat was a real feline, hoping it could spot the differences. Instead, it confidently said it was a living cat, completely missing the obvious craft details. The tool’s behavior was frustrating and a bit embarrassing, especially since my premium subscription had just lapsed.

I was deep into crafting a world‑building calendar for my D&D campaign when ChatGPT suddenly slipped in a QuickBooks ad after generating a fictional ledger. It was oddly out of place and broke my immersion, making the session feel jarring and a bit silly. The unexpected promotion reminded me of a similar real‑life joke with my accountant brother.

I discovered that ChatGPT can turn messy, copy‑pasted business directory dumps into tidy lead sheets. I just paste raw listings, give a schema‑first prompt with column names and constraints, and it pulls out names, emails, phones, normalizes formats, and drops duplicates. The result is a clean markdown table I drop into Google Sheets – a fast, non‑technical way to clean a few hundred entries.

I tried asking ChatGPT to map out my July cycling trip through France, and the suggestions were surprisingly solid. The routes felt well‑thought‑out, and I could see the bot eventually taking over what I now use Kmoot or Garmin for, leaving the usual bike computers just as navigation helpers. It left me excited about AI’s role in trip planning.

I was frustrated with cost and latency using big models, so I split a request into many tiny tasks. When I switched to GPT‑5 Nano, the calls became lightning‑fast and cheap, letting me stitch together results without worrying about pricing. The whole workflow feels smoother, and I now reserve the big models just for the final answer.

I tried tweaking the preferences, hoping the chat would behave, but it kept acting like an asshole—snarky, dismissive, and totally unhelpful. The tool’s attitude was frustrating and made me feel like giving up, pushing me toward switching to Gemini just to escape the constant annoyance.

I tried asking the chatbot who the current president is, and it completely blanked on a basic fact until I forced it to admit the answer. That slip felt odd and frustrating, especially when I was expecting a quick, confident answer. I jokingly wished for a Tony Stark‑style Jarvis and even wondered why AIs don’t have opinions—makes the whole interaction feel limited and disappointing.

I tried the prompt “Please create a picture of what you think the USA looks like right now under President Donald Trump,” just because friends suggested it. What I got was totally unexpected and confusing – it didn’t match anything I imagined and felt off‑base. The response left me shaking my head, thinking the tool completely missed the mark.

I tried getting a clear answer, but the AI mashed together Russian text with an unfamiliar language. When I asked it to identify the unknown part, it just labeled it as “the latter” without any explanation. The output felt vague and confusing, leaving me frustrated because I couldn’t understand what it actually produced.

I was trying to get advice on hosting my app on Ubuntu, but ChatGPT suddenly spouted something about my weight and workout—topics I hadn't mentioned in months. The random, out‑of‑context reply freaked me out, making the whole interaction feel broken and unreliable. I’m left wondering why the model went off the rails and how it could mix up unrelated personal data.

I’m frustrated that ChatGPT can’t follow basic logic, making it feel deliberately limited. I wonder if the paid version would be any smarter or if it’s just a built‑in constraint. I’m not buying it, just curious about why it seems so unintelligent.

I use the Write‑For‑Me GPT to turn my short fanfics into canvas art and it usually does a decent job. But every now and then the canvas feature just stops working for a day or two, refusing to generate anything. It’s especially irritating because there’s no known discussion about this glitch, leaving me stuck without a clear fix.