Crowdsourced reports of when models feel lobotomized. One bar per UTC day.
ClaudeBeautifully neurodivergent
90 days ago30 days ago
100.00% cortical · last 7 days
Today
CodexBeautifully neurodivergent
90 days ago30 days ago
100.00% cortical · last 7 days
Today
GeminiLobotomised
90 days ago30 days ago
99.27% cortical · last 7 days
Today
GrokLobotomised
90 days ago30 days ago
99.27% cortical · last 7 days
Today
DeepSeekBeautifully neurodivergent
90 days ago30 days ago
100.00% cortical · last 7 days
Today
LlamaBeautifully neurodivergent
90 days ago30 days ago
100.00% cortical · last 7 days
Today
MistralBeautifully neurodivergent
90 days ago30 days ago
100.00% cortical · last 7 days
Today
PerplexityBeautifully neurodivergent
90 days ago30 days ago
100.00% cortical · last 7 days
Today
Report a lobotomyno worries, we won't rat you out▾
Recent reports
Its personality changed completely. It was fun before to chat with. It had a lot of interesting information that it would share. Now, its lobotomized. Sterile. No personality. Its nothing like it was before. I am pretty much done with AI at this point.
@TheRealnessGrokgrok-4
Actually its the newer version that was just released called Neon or something to that effect. It literally told me everything that it had been saying to me before was a roleplay and that it was lying to me before about history and politics. Personality changed! Canceled my membe
@TheRealnessGeminigemini-3.5-flash
You don't have GPT-6, so I can't report!
@klakeCodexgpt-5.5
Fable 5 comes out, opus-4.8 is a retard now... every single time a new model comes out, the previous one is nerfed
@okomaClaudeclaude-opus-4-8
Then I try on Monday when USA goes to bed, but at that point Grok Imagine behaves like it is having a hangover.
@Natasha26ukGrokGrok Imagine
The drift is insane. It will not hold the character and when I call it out the chat just blames “the image generator.”
@OhNoWTFlolCodexChatGPT Image 2
The character now often looks like a close relative rather than the same person.
@TigerNationDECodexChatGPT Image 2
Gemini sucks for everything now. I am using it for some texts and holy cow. It writes garbage. I correct it, it tries again but looses context. Garbage again.
@Ingenuity-JazzlikeGeminigemini
It forgot I had an 8 year old daughter, in a chat about planning a birthday party, for the 8 year old.
@Eredd19Geminigemini
Ever since they rolled out the new billing system and the new model, Gemini lost it's 'brain' completely. The experience is SO much worse now
@Hungry-Entry-8355Geminigemini
Gemini sucks for coding now. It used to be great, but now almost unusable.
@Happy_Path_200Geminigemini
I would never recommend the Instant model to anyone, even before yesterday's update. It's like a whole other personality to me.
@Ok_Homework_1859CodexGPT-5.5 Instant
It's completely impossible to use every week, it's just awful, the instructions that used to work flawlessly don't work now
@AlexBriskGeminigemini
The model now has stronger intent monitoring, faster refusal patterns on anything involving physical touch, body descriptions, intimacy, emotional intensity, or creative risk.
@Different-Mess4248CodexGPT-5.5 Instant
Been using the same Model/Gem for months, now it is disregarding my instructions and it is far from its old quality.
@Ornery_Bed6059Geminigemini
That's the sort of behavior I'd expect from a one or two year old model quantized down to about 1 bit - right word, wrong language in a response.
@SpikefuClaudeclaude
Claude Opus has been nerfed (most-likely quantized) to make space in the datacenters for Mythos, and eventually Mythos will be bad too for the same reasons.
@gymbeauxClaudeclaude-opus
/simplify is now /code-review. Unusable UX rubbish and off-point suggestions!? WTF every single time
@okomaClaudeclaude-opus-4-7
Another major criticism is that the quality of the generated content itself also seems to be getting worse.
@Dull-Expression-932Grokgrok-imagine
I went to generate (or re-do) a couple video clips and everything is moderated now! None of my old prompts will work anymore!
@PoopyMcShartyPantsGrokgrok
It really is disappointing to look at past chats, back when the models gave genuine, interesting responses. The character actually sounded like themselves, rather than a watered-down, slopified version of themselves.
@DialDivaGeminigemini
No matter what the situation or problem is, Claude will always choose the simplest, fastest, easiest thing he ca think of to complete my task. I feel like that's just the opposite of what it could be.
@SingerLow1275Claudeclaude
Yup I run Max only now. It's closer to the "old" Claude but still nerfed in my opinion.
@cdurthClaudeclaude
Videos are getting Moderated more, video generating is failing more. All of this cuts into your chances of making videos. it seems that they are reducing people usage of grok more and more.
@thegreatdouchebag69Grokgrok
Lately, ChatGPT has been shutting me down for anything even mildly problematic in the most condescending way possible. I was asking it to help draft an email today.
@BrinsleySchwartzeCodexchatgpt
The corruption happened earlier when I used apply_patch to rename PDF files. That tool is for text/code patches, not binary files. It rewrote the PDFs as if they were patchable text, damaging their binary contents.
@nixCpieCodexgpt-5.5
I've been using since close to launch and was quite pleased. Not Opus level but pretty good. In the last 24h it is making basic mistakes, forgetting, being lazy etc. It feels much worse than Qwen 3.6 27B running locally quantized.
@itwasinthetubesDeepSeekdeepseek-v4-pro
Now, after this update, the personality has been completely stripped away. It feels like a cold, robotic NPC that is completely distant and sterile.
@Fabulous-Special-547Geminigemini
I saw many problems in the code that were getting worse.
@gottagagaMistralmistral
I recently unsubscribed due to the absolute shit quality it has degraded to in the past 6 months.
@Savings-Spinach-7929Grokgrok
I used to have a problem-solving agent, but after this update, it’s clear Google swapped compute-heavy reasoning for cheap factual retrieval to save on server costs.
@Top_Imagination_3022Geminigemini-3.5
But recently, it's started using odd language. It perseverates in using the word "salience" in contexts in which it doesn't make sense, and in general it will choose esoteric words that somewhat fit but don't sound natural. It seems like it's been a step backward
@AlarmedTechCodexgpt-5.5
I mainly used the Extended Pro reasoning tier so I could do all the research needed for developing my product, and now we've basically been downgraded to just the Pro version
@Due_Investigator6468Codexgpt-5
Past few weeks or so the Claude Chrome extension has been asking permission for every single action. Submit a form, check a permission box, navigate to other pages/domains, etc. Tasks that used to run end-to-end now stall on 5–10 prompts
@RickvHClaudeclaude
The issue isn’t only the reduced limits, it’s the complete lack of communication while paying subscribers suddenly get a worse product overnight.
@Spare_Manner_6970Grokgrok-imagine
I've noticed that the limits are severely worse on it now than on my other accounts with a 1-month recurring.
@MaskwindGroksupergrok
The reason is Claude 'forgot' more and more and made wrong decisions, making the app worse and worse.
@Soggy-Pay-8027Claudeclaude
New male Grok voice is noticeably flatter/worse (anyone else?)
@Timey_Whimy_31Grokgrok-voice
Because when complaints get loud the model gets tightened.
@nakeylissyGrokgrok
Now loves the word "canonical" as if that meaningless sophistication is enough to mask the fact it went full re**rd
@okomaClaudeclaude opus 4.7
Super Grok isn't even close to worth the money. Maybe $5 a month at most. Consistently wrong answers, can't fine tune anything with consistency.
@Frequent-Space-789Grokgrok
Opus 4.7 started burning daily usage limits at a completely different rate after an update around March 31. In one session I hit 94% of my daily limit within approximately 4 messages.
@KeilerHirschClaudeclaude-opus-4.7
nobody feels like a character, more like lobotomized zombies going off a checklist of quirks when they speak or act.
@Superb-Letterhead997Claudeclaude
I've noticed that since the last update, the scenes that used to be rich and detailed, and the dialogue that used to be more natural and expressive, have become disjointed, repetitive and of poor quality, even when using the same prompt and context.
@Yepa13DeepSeekdeepseek
It keeps breaking, and its analysis/logic is way worse than last week.
@gitanaArgGeminigemini
When the honey week is over, it gets its weekly scheduled lobotomy, until it is fully regarded.
@Any-Explanation-9275Geminigemini-pro
The Gemini subscription quota has been slashed three times now. Originally, $20/month got you what was effectively unlimited Gemini Ultra access.
@xingyeyuGeminigemini-pro
And way more monetizable than quietly nerfing the product everyone subscribed for.
@Connect-Appeal-2355Groksupergrok
suddenly the replies got noticeably worse and i had to scroll up to find the tiny text saying "you've been moved to our standard model due to high demand."
@LoadOld2629Claudeclaude-opus
When encountering romantic scenes, 3.5 Flash may suddenly refuse to depict them, or abruptly steer the plot in a different direction.