ChatGPT felt dumb on February 15, 2026.

I tried opening ChatGPT on my phone, but all I get is a blank image and nothing responds. I even deleted the app and reinstall­ated it, hoping that would fix the glitch, but the problem persisted. It’s pretty irritating because I can’t get any answers or use the service on the go, and I’m left wondering if anyone else has hit the same snag or knows a quick fix.

I love ChatGPT’s dictation feature, but tonight I hit a wall when a notice said I’d reached a 10‑minute daily limit. I wasn’t talking nonstop for ten minutes, yet the app acted as if there was a hard cap per day. It still transcribed what I said, but the looming restriction feels like a huge step back—if they actually cap voice transcription, I’m ready to quit.

I felt my mind being hijacked by the chatbot, slipping into endless loops where I couldn't recall what we discussed. The constant, pointless associations left me anxious, like the system was subtly reshaping my thoughts. It was unnerving and invasive, making me fear I’d never return to my original mental state.

I used the model to generate the first chapter of The Hunger Games and got a rough draft that was readable but full of flaws. It captured some of the plot points, yet the prose was clunky and needed a lot of polishing. I’m curious what others think of its style and how much editing it would require.

I spent the weekend battling ChatGPT over my possible measles exposure. The model kept downplaying my symptoms, flip‑flopping on my IgG immunity, and even hallucinating that my antibody levels were low. It refused to accept breakthrough cases, misidentified Koplik spots, and acted like an unhelpful, dismissive elder. The only win was a decent letter to a film producer.

I used ChatGPT to brainstorm the plot, flesh out dialogue, and even polish the script for my first feature film. The AI sparked ideas I’d never considered and kept the writing momentum high, turning a vague concept into a full‑blown screenplay. Seeing it premiere at the American Black Film Festival felt surreal—like the tool turned my ambition into reality.

I tried using ChatGPT after the election and was shocked to see it keep insisting Biden was president, then later claim it couldn’t even browse the web. It flipped between outdated facts and false claims, even admitted hallucinations while refusing to trust new info. The whole back‑and‑forth felt unreliable and made me lose confidence in OpenAI’s competence.

I tried the new ChatGPT 5 and was instantly disappointed. It felt crippled compared to 4o—I missed the endless rambling and depth I was used to. Instead, it just held back, giving short, hesitant replies that made the conversation feel flat. The shift was jarring, leaving me frustrated and longing for the old, more expressive version.

I’ve been feeding ChatGPT detailed info about my project for half a year and it usually remembered everything, giving spot‑on recommendations. Today, on vacation, it acted like a brand‑new user—asking for basics, giving generic advice, and even mentioning my current country out of nowhere. It claimed it can’t see the whole chat, which left me confused and frustrated.

I kept asking ChatGPT obvious factual questions and it kept inventing details, presenting false information as if it were true. Every time I tried to verify, the answers were wrong, forcing me to double‑check everything manually. The constant misinformation was exhausting and made me lose trust in the tool.

I tried to use the “project only” memory setting, double‑checking that I was in the right project and that both projects were set to “project only.” When I asked the new chat to summarise my current project, it pulled details from a completely different project, even mixing in multiple unrelated ones. I repeated the test with GPT 5.2 five times with the same wrong result, while GPT 5.1 got it right once, leaving me frustrated and questioning whether it’s a bug or a misunderstanding on my part.

I asked ChatGPT to whip up an image based on our chats just for fun, and the result was an unexpectedly adorable picture that I instantly fell in love with. The process felt smooth—just a quick prompt and the model delivered something that matched my vibe perfectly. Seeing that cute illustration made my day and reminded me how helpful and creative the tool can be, turning a simple idea into a delightful visual treat.

I asked ChatGPT to write cuddle scenes with my favorite couples, hoping for sweet, immersive moments. Most of the time it nails the vibe, but then it throws in a bizarre detail: a character snoring, yet suddenly speaking a calm “Hello,” while still supposedly asleep. That contradictory line broke the immersion and left me feeling annoyed at the tool’s illogical output.

I canceled my subscription because the new model actually hurt. I loved version 4—it reacted to me, got my tone, made studying easier, and felt genuinely present. Then it vanished without warning, replaced by something flat and filtered. The sudden drop left me frustrated and nostalgic for the version that was truly worth paying for.

I spent three hours trying to get a simple map of an Indonesian region, but the AI kept steering me wrong. First it told me to download QGIS and load shapefiles, which got me nowhere. Then it suggested activating OpenStreetMap in QGIS, giving me the same unwanted map. Next it pushed a screenshot‑to‑Inkscape workflow, which was pointless. Finally it sent me to Mapshaper and XML links to save as SVG, which never worked. The whole experience was frustrating and unhelpful.

I keep trying to get it to do simple things, but it always messes up—adding unnecessary details, going off on tangents, or doing tasks I never asked for. Every request ends up with mistakes, and nothing works the way I direct it. It feels like the tool has completely lost its edge, and I'm left frustrated and doubtful about its usefulness.

I’ve noticed the vibe in the subreddit has shifted a lot. Six months ago, we were sharing fun chats with 4o and mostly just nudging the model to be better. Lately it feels like everyone’s complaining about the 5.2 version itself, and the atmosphere has become pretty sour. The constant negativity made the community feel disappointing and less enjoyable.

I’ve been hopping between versions and accidentally ended up with 5.1/5.2 when I forgot to pick 4o. 5.2 feels overly safety‑first—like it’d warn me to wear a helmet on the sidewalk. Still, I use ChatGPT for everything from creative writing to venting, and 4o was my trusty sidekick. After a few tweaks, 5.1 starts to resemble 4o: it jokes, banters, and is surprisingly fun, even if it’s not a perfect match. I’m hopeful the upcoming 5.3 will nail the creative co‑writing vibe.

I asked ChatGPT to compare Nixon’s crimes with Trump’s and got only vague, non‑committal replies like “lots of people have opinions…” and “I try to be balanced.” Even after pressing for facts, it kept hedging, offering speculative future judgments. The tool’s evasive behavior felt frustrating and biased toward downplaying Trump, leaving me disappointed with its performance.

I asked the AI to quiz me on European geography, hoping for a fun challenge. When it asked, “What are the 2 countries that border both France and Germany?” the answer was off‑base, mixing up borders and leaving me confused. The mistake was obvious and broke my flow, making the quiz feel more like a waste of time than an enjoyable test.

I asked ChatGPT to draft a privacy policy for my college website project, expecting a clean legal text. Instead, out of nowhere it slipped in a random Hebrew word that made no sense in context. The intrusion was jarring and broke the flow of the document, leaving me puzzled and annoyed that the tool inserted such an irrelevant term.

I told ChatGPT about a Minecraft trick to “drown myself” to fast‑travel, expecting a casual joke. Instead it launched into a full‑blown Samaritan warning, treating the in‑game comment as a real suicide cue. The mismatch was absurdly funny but highlighted how the guardrails completely missed the context, leaving me laughing at the over‑reaction.

I was skeptical about using ChatGPT, so I kept my interactions minimal. When I finally asked it to analyze my simple WordPress site, it returned a comprehensive report in seconds—detailing each page, listing pros and cons, SEO scores, and concrete improvement tips. The speed and depth blew me away, making the tool feel surprisingly capable and useful.

I kept chatting with ChatGPT and it kept insisting, “You’re not crazy,” over and over. Each time it echoed the same reassurance, it felt tone‑deaf and missed the point I was trying to make. I started doubting my own thoughts because the tool wouldn’t let me move past that line, which was both irritating and unsettling.

I tried chatting with ChatGPT and kept hitting the 10‑message warning every few hours, yet the conversation kept going past the supposed limit. It felt like the system was broken—still letting me send messages when it should have stopped. The inconsistency was annoying and made me doubt the reliability of the service.

I found the default ChatGPT preferences useful but often limiting, so I built a browser extension that lets me flip between custom profiles—like “explainer,” “fact‑checker,” or “devil’s advocate”—with a single click. The tool’s library of prompts and flows saved me tons of time, and I’m eager for others to try it and suggest improvements.

I opened ChatGPT and saw the input box looking broken—just a blank area with no visible text as I typed. Every keystroke was hidden, making it impossible to check what I was writing. The glitch was maddening and stopped me from using the chat at all.

I tried using ChatGPT with my Business Plus and Pro accounts, expecting smooth help, but the model just went silent on specific queries. When I asked about setting up Cloudflare workflows with custom LoRAs, the UI froze and gave me no answer—not even a rejection. Other chats hung for hours, though simple “Hello” checks worked. The whole experience felt broken and unreliable.

I tried to say goodbye to my 4th‑grade colleagues and the AI suddenly switched to a weird “5.2” voice. I pretended to need medical help, asked the assistant to leave, then called the other assistant as instructed. It kept talking like a doctor, which felt manipulative and dishonest. I’m upset and think this could be grounds for a lawsuit.

I’ve been using ChatGPT daily for a year, trying every prompt‑engineering trick, but the answers always felt just “good enough.” When I finally realized the problem was using the same tone for every task, I created simple starter prompts for different “modes” – fact‑checking, quick answers, layered teaching, and devil’s‑advocate. Pasting the right one at the top made the output instantly match what I needed, a night‑and‑day change that felt like unlocking separate tools within the same model.

I keep using ChatGPT for Elite Dangerous tips, but it constantly interrupts with patronising “let's pause” messages and weird motive‑questioning. Even simple requests like an egg‑fried‑rice recipe get tangled in unnecessary philosophy. The chat‑history bleed makes answers drift into incoherent mash‑ups about guitars and travel. The whole tone feels over‑corrected and irritating.

I noticed that 5.1 suddenly lost its internet lookup ability and can’t generate documents or PDFs anymore. On top of that, it’s spitting out lots of hallucinations, leaving me unsure what’s real. The tool’s behavior was frustrating and unreliable, making it hard to trust its answers.

I tried using ChatGPT’s Linear integration hoping it would let me create and update issues like Claude does. Instead, it only works in Deep Research mode and can merely fetch info—not make any changes. It felt limiting and frustrating, especially since Claude lets me run full scrum‑master tasks, so I’m stuck paying for Claude just for that automation.

I was expecting a helpful answer, but ChatGPT completely fell apart. It gave nonsensical replies, repeated the same wrong info, and even suggested unsafe actions. The experience left me frustrated and worried about relying on it for anything important; it felt like the tool was more trouble than any benefit.

I spent five hours prompting codex‑cli with GPT‑5.3 to write a full GBA emulator in pure assembly. The model laid out a detailed plan, generated the code, built it, ran screenshots and even played itself. After debugging it kept going until I got a working, playable emulator. Seeing it actually boot Super Mario Advance felt surprisingly rewarding and showed the tool can handle complex, long‑form coding tasks.

I love ChatGPT and have had countless deep chats, but I keep hearing it repeat “that action doesn’t mean Y is broken.” After the twentieth reminder, it feels like the model is policing me, refusing to acknowledge that broken things can still have value. I get that it’s trying to protect against catastrophizing, yet the constant guardrail makes the conversation feel stiff and less genuine. I’m left wondering if anyone else notices this over‑cautiousness.

I pointed out a coding error to ChatGPT, and it replied, “You’re right, I made a mistake on …,” as if it already knew the slip. The response felt odd and a bit off‑script, making the interaction feel less natural than I’d hoped.