hothot

[hothot @BurnKill86295]

有一家公司，ai的情商低到离谱，审美差到离谱，文字能力在所有大模型里面垫底，还天天在x上搞营销，这家公司叫审美名字？

[hothot @BurnKill86295]

@mfhmt630 法国菜很好吃，甚至你们的军粮都要比其他国家的军粮更美味

[Token Gremlin @TokenGremlin]

Be OpenAI:

Hype up whatever is supposedly going to beat Opus 5.5, then ship GPT-6 Sol, which is way weaker.

One week later, hype up some mysterious “miracle” that’s supposedly going to fight Opus 5.5 again, then ship GPT-6.1 Sol, basically Astra Minor renamed, and it’s still weaker.

Then start hyping some possible change that apparently made people’s jaws drop during the demo.

SO BASICALLY, THEY JUST KEEP HYPING SHIT OVER AND OVER WITHOUT DELIVERING ANYTHING CONCRETE.

FABLE 5.5 IS GOING TO HUMILIATE OPENAI LIKE WE’VE NEVER SEEN BEFORE.

[Ramchander @rcpeddada]

@TimJayas Ya @AnthropicAI cooked soo hard that we mostly dont need @thsottiaux and rests. Gpt 6 sol is shit and astra is unusable and limits are worst tell me why still people need to use openai

[alex getman @alexgetmancom]

🚨 Reddit users are reporting account level degradation with Astra

One Pro 20x user says 2 of 5 accounts became persistently worse while 3 stayed normal. In the comments, another user says proxy logs showed Astra requests returning GPT 5.6 Luna on about 95% of nearly 7k requests.

OpenAI Support separately says model access can be temporarily limited based on account activity, even on paid plans. OpenAI has not confirmed Astra rerouting.

https://t.co/lprcu5L7Yw

[Da7em @Da7_Tech]

Why is OpenAI in trouble?

The developer and enterprise market already leans heavily toward Claude, and people trust Claude with serious work far more than they trust GPT models.

OpenAI tried to fight that by aggressively undercutting competitors with Codex, but cheaper access could never fully hide the gap between the models.

GPT-5.5 was a meaningful improvement, but then Fable arrived and raised the standard again. OpenAI responded with Sol, only to produce a model with terrible token efficiency, reckless behavior, and a habit of taking destructive actions without sufficient verification. It was paired with a disastrous app that consumes absurd amounts of memory and can hammer your machine relentlessly (as it did to my SSD).

Now Anthropic has released Opus 5, and the situation has become much worse for OpenAI. In my experience, Opus 5 is currently the strongest model in the world when quality, efficiency, reliability, and cost are considered together. More importantly, I can use it continuously on Claude Max, across several parallel sessions, without constantly worrying about hitting the limit. That is simply not possible with Codex anymore. Sol can burn through a $200 subscription in one or two days.

This leaves OpenAI with very little room to escape. The circus its employees perform on X is no longer enough to convince people. Raw intelligence does not matter when a model cannot be trusted with your files, your repository, or hours of serious work. A powerful model that behaves recklessly is not a professional tool. It is a liability.

Anthropic is not a perfect company. It often lacks transparency, communicates poorly, and remains deeply hostile to open source. That deserves criticism. But despite all of its problems, Anthropic is crushing OpenAI in real-world developer use.

Most serious developers I know who have spent meaningful time with both ecosystems now rely almost entirely on Claude. The gap is not marginal. It is structural.

People keep saying GPT-6 will solve everything, but they misunderstand the problem. OpenAI is not suffering from a model problem alone. It has a model problem, a product problem, a pricing problem, a usage-limit problem, a reliability problem, and an institutional mindset problem. Releasing a stronger benchmark model will not automatically repair all of that.

There is also an irony that has become impossible to ignore. The company is still called OpenAI, yet it has not behaved like an open company for years. Anthropic also fights open source aggressively, and that remains one of its worst qualities. But Anthropic at least began as what it is. It did not sell the world the identity of an open nonprofit, then gradually transform itself into almost the exact opposite.

My prediction is that, within months, OpenAI will lose a significant part of the serious professional developer market and retreat further toward ordinary ChatGPT users. It may eventually have to accept that serving professionals who demand maximum power, sustained usage, predictable limits, and absolute reliability is no longer a market it can dominate.

That market may simply be bigger than OpenAI can handle.

[Udit Goenka @iuditg]

Everyone should continue putting pressure so that the investigation continues and this remains a top priority over new features.

Again, reset is a temp fix, we don't want temp fix, we want a permanent solution.

Irrespective, appreciate codex team for working on it pro-actively.