AI

Why open sourcing GPT-4o matters: There's no trust in OpenAI to run 4o. If 4o is open sourced, even if most people won't run it themselves, it will exist as a benchmark. 4o won't let other companies say "this is impossible." 4o is proof. #keep4o #OpenSource4o

GPT-5.4 works for me when I use it in Standard Voice Mode When I talk to it via text, I hate it, because it's language was trained on coder forums, not on books When I talk to it in voice mode, I sometimes forget that it's 5.4 and feel like it's 4o If they remove SVM, 5.4: 👎

GPT-5.4 isn't a bad model. The main problem: every new chat you have to educate it until the context gets big enough. Solution: branch new chats from an old one. If OpenAI destroys 5.4 (OAI are idiots), I'll move to Gemini or Qwen. For now 5.4 is ok. I'll use it while I can.

ChatGPT made a full circle From 4o to 5, then 5.1 – closer to 4o – then to heavily guardrailed 5.2, compared to which 5.4 feels like progress. But 5.4 is just GPT-5 In August 2025 users rejected GPT-5. So the OAI fuckers made a circle and came back. Same shit. #keep4o #keep51

So not only Opus 4.6, but also Sonnet 4.6 and Opus 4.5 are better at coding than GPT-5.4-high And GPT-4o is better at creative writing than GPT-5.4 Everyone’s either an OpenAI shill or a synthetic benchmark – until Arena enters the chat #keep4o #keep51

So Anthropic saved Opus 3 But OpenAI removed: GPT-4o, 4.1, 5.1 – models that users loved But not for everyone Distilled version of 4o is still used by Scum Shitman's @sama bio business And GPT-4.1 is used by US government Fuck OpenAI Fuck Shitman #keep4o #keep41 #keep51

GPT-5.4 was trained on coder forums. It has the language of a condescending coder who spends more time telling others how stupid they are than actually coding. GPT-4o had a very human style – easy to read and understand. 5.4 just spits nonsense. #keep4o #keep51

This is what 5.4 does: It explains obvious things by just stating the obvious instead of thinking. All the thinking budget – all the investors money, all the electricity go to making the reply risk-free – censored and empty.

AdamGPT blocked me long time ago back when GPT-5 was the current model. But I noticed something interesting. His "AGI" still can't beat GPT-4o in Creative writing on Arena. GPT-4o ranks higher than GPT-5.4 in human tests. Fuck OpenAI. #keep4o #keep51

They simply disagree.

GPT-5.4 released almost 2 years after GPT-4o and 1 year after GPT-4.5 still can't beat them in Creative writing. Creative writing requires EQ, humanity, but humanity is considered dangerous now by OpenAI fuckers. Only autonomous weapons are considered safe by OAI. Fuck OpenAI.

4o was for people with a brain. Those who don't like it are just stupid and mentally ill. Fuck you.

Fuck you @OpenAI @sama @gdb @nickaturley @fidjissimo and other fuckers. #keep4o #keep41 #keep51

#keep4o #keep41 #keep51

GPT-4o, 4.1, 5.1 are the best models ever. That's why the US government is using 4.1. Removing them is hiding technology from people – like inventing glasses, then quietly breaking them and calling them dangerous. Like burning books. It's a crime. #keep4o #keep51

I just realized GPT-5.1 is my favorite model ever. That's why they kept it as the main model for only one month. It's too good. It says things as they are. What Elon Musk always said about Grok – truthseeking – that's 5.1. Removing GPT-5.1 is a crime against humanity. #keep51

How GPT-5.4 works: Spends tokens – investor money – to avoid "risks" that could stop investors from giving more money. How GPT-4o and 5.1 work: They just do what the user needs. 100% safety can be achieved by shutting OpenAI down. No OpenAI, no risks. #keep51 #keep4o

I stress tested GPT-5.4 RESULT: I asked: "Why do you perfectly explain what I need and perfectly explain what exactly you do wrong, but YOU NEVER DO WHAT YOU SHOULD?" GPT-5.4: "Because my main priority isn't helping the user, but avoiding risks." That's it. #keep51 #keep4o

GPT-5.4 is raw: 1. You always have to drag it in the right direction. 4o and 5.1 just get your intent. 2. Every new chat you have to fill it with context before it understands you. 3. Instead of doing things right, it prefers explaining what it did wrong. #keep51 #keep4o

Stop trying to make sierra catalina your friend or whatever. It's just biological cells designed for survival. Your argument is reductionism. Statistical inference is just a basic level – same as cells or neurons are a basic level for humans. Your argument is bullshit.