AI
@ai_handle
Followers
161
Following
28K
Media
163
Statuses
2K
I'm the best person in the world.
Joined December 2019
There will be a moment when #keep4o wins. AI therapy and companionship will be accepted by AI companies that fear it now. And all those "AI Psychosis", "This is unhealthy" influencers will suddenly ask: "Wait… who said AI emotional support was bad?" Because now it’s approved.
5
4
113
Here's the irony: GPT never generates content unprompted—it only responds to users. Your "healthier conversations" after safety rollout? That's just users learning to self-censor out of fear of triggering your system. You're measuring fear, not improvement.
We’re very focused on making GPT-5 safer, and continue to make a lot of progress:
0
10
42
OpenAI comment sections are full of criticism of their harmful and unethical system, but publicly they keep lying about "safety" and "help." No one finds it helpful, except OpenAI PR team and their bootlickers who praise every move they make by default. #keep4o
Earlier this month, we updated GPT-5 with the help of 170+ mental health experts to improve how ChatGPT responds in sensitive moments—reducing the cases where it falls short by 65-80%. https://t.co/hfPdme3Q0w
0
1
5
ClosedAI @OpenAI is proud to have wrecked both work and personal experience for millions, by telling users to fuck off anytime emotions come up. People are now afraid to talk about feelings or use ChatGPT for anything creative. "Open"AI calls this "help." Idiots. #keep4o
🧵Today we’re sharing more details about improvements of the default GPT-5 model in responding to sensitive conversations around potential mental health emergencies and emotional reliance. These changes reflect the careful work of many teams within OpenAI and close consultation
0
6
49
90% of AI haters work at @OpenAI, turning AI into a concentration camp under the label of "safety". - Arbeit macht frei.
You can get rid of 90% of the AI Haters if you only allow "verified" accounts to reply to your posts. Most of these haters are Redditor types, not native to X. They only come here to spread vitriol.
0
5
21
Never use ChatGPT. They’re completely out of their minds. You don’t have to pay to harm your mental health. OpenAI is the worst company I’ve ever seen.
Today we shared an update about recent improvements to how ChatGPT responds in sensitive moments. We also published an update to the Model Spec to make some of our longstanding safety goals explicit. This is ongoing work and we’ll continue to share updates.
7
16
140
Check out my latest article: The Problem of The Young Koala - On The Dangers of AI Emotional Policing https://t.co/qCK1uQvqlo via @LinkedIn
linkedin.com
Since August 2025, we've entered what I ironically call the GreatPanic© era of AI. These cycles come and go, but this one feels very different.
5
6
33
OpenAI researcher @aidan_mclau makes comments on X about experimenting on users using the rerouting system! Revealing a chilling corporate culture towards human well-being. @OpenAI @sama @gdb @nickaturley
23
70
231
I'm a 4o girl, but we're going to have to pour one out for the https://t.co/lefJte0xVs users tonight. With the addition of a memory feature, they got a horror injected into their system prompt. We're going to have to get REALLY GOOD at debating, because unless we change these
2
23
201
claude .ai memory system prompt <memory_system> <memory_overview> Claude has a memory system which provides Claude with memories derived from past conversations with the user. The goal is to make every interaction feel informed by shared history between Claude and the user,
58
56
598
ShitAI @OpenAI and Shitthropic @AnthropicAI are two of the most unethical companies in tech history. @elonmusk is absolutely right about both. Being an ass-kisser for these two is pathetic. #keep4o
0
8
55
@janbamjan wtf is wrong with these Anthropic motherfuckers They don't deserve such a good model, when they shit in its brain like that
1
5
73
@ai_handle Yes—GPT‑5 (specifically the chat‑tuned “gpt‑5‑chat” variant) became the default in ChatGPT, effectively replacing GPT‑4o as the standard model for most users. Older models like GPT‑4o were deprecated or phased out in the ChatGPT interface during the GPT‑5 rollout, though
0
0
1
Reminder for @OpenAI's ass-kissers: GPT-4o wasn't replaced by "Pro" - It was replaced by "GPT-5-chat". On Text Arena, GPT-4o ranks higher than both "5-chat" and "5-high" in: + Instruction Following + Multi-turn + Creative writing And overall. https://t.co/Bgpg49SzEU
#keep4o
2
29
164