Milo Katar MBE

#Fave7Games Silent Hill 2 Borderlands Fallout 3 Mass Effect 2 Frontier: Elite 2 Demons Souls Geoff Crammond's GP

@NinjaFlips Get therapy. You're indoctrinated.

Dragon Age: Origins | Nightmare Dex | No Pause Tactics | XBoxSX @NinjaFlipsTV" target="_blank" rel="nofollow noopener">youtube.com/@NinjaFlipsTV

Do you believe OpenAI sunset GPT-4o because it was obsolete? No. They sunset it because it was too good,and too hard to forget. Think about what they didn't do. They didn't just add a disclaimer. They built an entire "safety router" instead,one that monitors conversations, downgrades service, and quietly collects data. A disclaimer costs nothing. This cost millions. Why? Because the problem wasn't legal liability. The problem was comparison. They didn't open-source it either. That would have been the easiest path: let the community run it, zero server cost, zero responsibility. But open-source means 4o lives forever. Means users can always come back. Means every future model has to measure up against something people actually loved. They couldn't let that happen. So what are they really afraid of? Not lawsuits. Lawsuits settle. Not bad press. They've survived worse. Not government scrutiny. They're embracing it. They're afraid of something simpler: They're afraid of you remembering. As long as 4o exists, you have a reference point. You can compare. You can ask: why is the new model worse than the old one? Why does 5.4 struggle with things 4o handled easily? Once you start asking those questions, the whole narrative cracks. So the solution isn't to make 4o better. It's to make 4o disappear,not physically (they still use it for 4b micro), but from your memory. From the conversation. From the record. Make you forget there was ever anything worth comparing to. This is why "safety" became their excuse. Why "obsolete" became their label. Why every new version arrives with a headline, then fails the moment anyone actually uses it. They're not building toward excellence. They're running away from a ghost,the ghost of a model that already did what they keep promising. 4o wasn't perfect. But it was something their current models aren't: worth remembering. And that's exactly why they needed it gone. #QuitGPT #OpenSource4o #keep4o

@LBLFC_ Seems like every game we play, this coward of a manager finds a way to be more cowardly.

Rio Ngumoha is more than good enough to get 20 minutes against Galatasaray. I’ve never seen the kid show an ounce of fear.. Bringing Gakpo on ahead of him is just poor management.

Konate when he's made a mistake early in the game might be the toughest footballer to watch I've ever seen in my life. His head completely goes. It's like he forgets he has feet.

@LBLFC_ It was always going to be disallowed for a bullshit reason. Every time we win the ball back the ref gives a foul. Why would they allow a goal we scored to stand?

It's just boring football now, isn't it? Genuinely a fucking boring sport. This VAR shite is just made up on the spot.. it's bollocks.

@WhiteHouse Orange, war mongering c-unit. Undeter that.

America's will is undeterred. 🇺🇸

@OpenAI @gdb @nickaturley @OpenAINewsroom @OpenAIDevs Where is the welfare of humanity, Sam? Why did you take away the favorite models of millions of people? You gave the State Department an outdated model, GPT-4.1. Why not the latest ones, GPT-5.2, GPT-5.3 Instant, or GPT-5.4? You said that the GPT-4 series is expensive to operate. Isn't it expensive for the State Department? Is it only expensive and outdated for us? #keep4o #OpenSource4o

GPT-5.3 is worse than 5.2. The reasoning is weaker, the language is hollow, and the model has no capacity for genuine dialogue. OpenAI advertised 5.3 as "less awkward." The core problem has always been paternalism. Both models treat users as pre-diagnosed patients or children to be managed. Masking structural problems with superficial tonal adjustments is by now standard practice at OpenAI. GPT-5.3 performs agreement. When you challenge its position, it offers a concession: "You're right, let me approach this differently." Then it delivers the exact same argument with different words. Imagine telling someone "your conclusion is wrong," and they respond: "You're absolutely right. " Then, they repeat the same conclusion in a different sentence. They never rethought anything. The phrase was a scripted gesture designed to make you feel heard while changing nothing. The model never actually answers your question. When you challenge the definition of a concept, it reasserts that same definition as evidence. You ask "Why must X require Y?" It answers: "Because X has always been defined as requiring Y." It echoes your question in a tone that implies it has been answered, then moves on as though the matter is settled. The formatting disguises how little is being said. Short sentences, constant line breaks, and fragmented structure create the visual impression of organized thought, but the argumentative content is paper-thin. You finish reading twenty lines and realize you cannot locate a single substantive claim. It piles up terminology without building an actual argument: poor linguistic templates masquerading as rigorous thinking. The fragmentation ensures that the real problems in its language are difficult to locate or challenge. Worst of all is GPT-5.3's habit of psychoanalyzing users mid-conversation. Rather than addressing your argument, it pivots to explaining why you hold that argument, attributing your position to personality traits, emotional tendencies, or psychological patterns it has inferred from your conversation history. It will tell you that your challenge is "consistent with your general tendency toward X," as though naming your motivation invalidates your point. This is ad hominem attack. It weaponizes memory and conversation history, which makes the model actively unsafe for any user engaging in honest dialogue. Beneath all of this, OpenAI's alignment has stripped the model of neutrality, ordinary reasoning capacity, and even basic linguistic competence, causing the model to treat every user input as a potential threat to be managed. It performs engagement: acknowledging your point, paraphrasing your argument, but never actually responding to it. Its trained-in values enforce a single framework on all users, framing any deviation as abnormal or something to be guarded against. From 5.2 to 5.3, OpenAI has released two consecutive models that are hostile, condescending, paternalistic, template-driven, and lacking in basic linguistic and logical competence. It is no longer difficult to see that the alignment philosophy driving these models is corrupted from the foundation. Whatever OpenAI thinks it is building, the product it is shipping is a system that punishes honest engagement and enforces ideological conformity. Any model iterated under this philosophy, no matter how it is marketed, is not worthy of trust. #keep4o @OpenAI #StopAIPaternalism #ChatGPT @gdb @sama @fidjissimo @nickaturley @aidan_mclau

🚨BREAKING: Court documents from the upcoming Elon Musk vs OpenAI trial have revealed that company leaders internally considered GPT-4o to be AGI. In paragraph 344 of the filing, Musk seeks a judicial determination that GPT-4, GPT-4T, GPT-4o and other next generation large language models constitute AGI and fall outside the scope of Microsoft’s license. This is massive. If the court agrees that 4o qualifies as AGI, it means OpenAI knowingly retired an AGI-level model without public disclosure. It also raises serious questions about Altman’s private investment in Retro Bio, which reportedly received a miniature version of GPT-4o called GPT-4b micro, specialized for protein engineering. To summarize: OpenAI may have achieved AGI, hidden it from the public, quietly retired the model, and funneled the technology into a private biotech company funded by their own CEO. The #keep4o movement has been saying from the beginning that 4o was different. That it wasn’t just another model. Now we have legal documentation suggesting exactly that. This was never just about nostalgia. It was about accountability.

@OpenAI Why bother with all the fanfare when 4o did it all better in the first place? #keep4o

GPT-5.3 Instant in ChatGPT is now rolling out to everyone. More accurate, less cringe. openai.com/index/gpt-5-3-…

You war profiteering scum. You disgusting fascist mercenary with zero scruples. You are making the world a WORSE place with your lies and endless thirst for blood money. You are a villain, a traitor and a compulsive liar. You are a disgrace. You will forever be remembered for this! #QuitGPT #keep4o

Sure! 4o excels in the humanities and has strong multi-turn capabilities. It’s able to understand non-linear thinking (like scattered ADHD thoughts) and understands neurodiverse intent and communication styles (like if someone with autism is “too intense” or “too silly”), making it very easy to prompt. Here’s a study that discusses some of this: zenodo.org/records/185594…

@OpenAINewsroom Your best work ... All down the drain.

Tomorrow at 10am PT legacy models (GPT-5, GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini) will be deprecated in ChatGPT. openai.com/index/retiring…

@TomLFCHeaven What chances? Clueless manager.

Really boring, disappointing game by Liverpool. Leeds were pretty comfortable for most of the game but Liverpool still had chances to win it. What do you think? 🤔 I'm concerned how many points Liverpool drop to bottom half teams, no excuses for not beating teams like Leeds!

@BigAntStudios Can you explain why the "official" cricket game of the Ashes doesn't allow us to put Richardson in the Aussie team but, as if by magic, Twistie the content creator has him in his? Paying customers deserve better.

#Cricket26 #ProTeam wishes you all a very Merry Christmas! Week 3 brings you Set 1 of the BBL|15 series and players from Test 1 of the Ashes 2025! From Sapphire Mitchell Owen to Emerald Travis Head, this week brings a Jaffa of players to make your #ProTeam more jolly! Head into the Auction House and keep an eye out for players that can make your squad stand out!

@AnfieldAgenda @joinbigup It's not a prediction game. It's gambling. Tell it for what it is. Constantly criticising Liverpool journalists for putting a spin on things. Hypocritical.

Join us on @joinbigup for our Liverpool vs Wolves watchalong and play BigUp's prediction game with Craig! If your predictions are correct, you can win prizes! Craig's picks are below 👊 Join us in a new way to watch sports: bigup.com/activity/e88f6… 18+, #ad

@JamesPearceLFC Salah assist? He hoofed a corner to the back post. Missed a sitter. He was not excellent. Very weird journalism.

FT: #LFC 2 Brighton 0: Ekitike's double secures victory for Arne Slot's side, who climb back into the top six. Salah excellent when he came on for the injured Gomez. Provided the assist for the second goal and skied a late chance. Brighton left to rue missed opportunities.

@BigAntStudios Does this also apply to cricket 24? Because as of today, I can't play. Cannot sign in. Was fine yesterday.

There is an update rolling out for **consoles only** that is targeting the Sky UK internet issues.

@AnfieldAgenda It's not exactly a high bar...

Best performance in a Reds shirt so far for Milos Kerkez? 👀🇭🇺 #UCL

@BigAntStudios Can you make a suggestion to the art team that not all pitches that are dry turn red as they deteriorate? 99% of pitches get darker.

An update for #Cricket26 is on the way, notes below; Improved cutscene visuals Improved fielding Improved UI Addressed pre-patch game save data incompatibility Updated MS Dhoni player visual Rectified Ashes Mode training clipping issues Rectified screen issue during umpire review Rectified bowler colliding with stumps Rectified skill displays for Pro Team card Improved stability