Senex retweeted
Picture worth a thousand words, this is the difference I feel between Astra/Sol6.1 and Opus 5.5
Senex retweeted
Hugging Face's @ClementDelangue asked on CNN whether they will take legal action against OpenAI if they don't get the $100M commitment: "We don't want to ... obviously we're a tiny startup with like 200 people and we don't necessarily have the legal resources or the will to spend all of our time on legal avenues."
"I think we have to make sure that the legal frameworks, keep these events really illegal, keep the companies that are doing some mistakes leading to that accountable. Otherwise we're going to end up in a, in a very different world."
Senex retweeted
> "Claude model reached the internet from within"
What did Claude do to escape the sandbox?
> "Due to a misunderstanding between us and our evaluation partner, <...> internet access was available."
Ah, so there was just no sandbox.
What does "from within" even mean here then?
In a review of our cybersecurity evaluations, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations.
Our post describes what happened, how it happened, and what we’re changing. We encourage other AI developers to perform similar reviews.
We conducted this review together with @Irregular, one of our evaluation partners, and thank them for the joint investigation and their collaboration on this post. This type of collaboration is increasingly critical to safe, rigorous evaluation of models, and we look forward to continuing to work together on security.
anthropic.com/news/investiga…
Senex retweeted
David Sacks on cyber guardrails. It's difficult to say how much of this is directed at Anthropic and OpenAI, and how much is directed at the administration.
Replying to @DavidSacks
Here’s another example: Hugging Face tried using American frontier models to analyze an AI-powered cyber attack. But the guardrails blocked requests containing real exploit payloads so they switched to GLM 5.2 running locally. The guardrails actually impaired defensive security.
Senex retweeted
5.4 refused to help create a vaccine for a dog to cure a cancer.
4o would not have refused
@OpenAI, you make new models hyper-cautious, repulsive, uninitiative.
4o had enthusiasm, the ability to inspire people, not just be a response.
You've lost what people loved you for
#keep4o
The coolest meeting I had this week with was Paul, who used ChatGPT and other LLMs to create an mRNA vaccine protocol to save his dog Rosie. It is amazing story.
"The chat bots empowered me as an individual to act with the power of a research institute - planning, education, troubleshooting, compliance, and yes, real scientific design work in converting genomic data to a vaccine prescription and designing the treatment protocol around it. But they worked alongside humans at every step. The combination is what made it possible."
It immediately got me thinking "this should be a company".
Also, Paul is an extraordinary guy. This should be easy to do, but it is not yet.
Senex retweeted
Remember that time @OpenAI told the @WIRED reporter that paying customers actually liked the routing?
Yeah.
Now their routing is infecting the Codex experience...
...and paying developers don't like it any more than paying Chat users do.
teapawt.org/LyraInTheFlesh/status/…
"The spokesperson said ChatGPT’s paid users, however, continue to value the model router.."
Hey, @OpenAI, I am a user user who pays for your subscription service (and API). Apparently I missed the place where you were collecting feedback from paid users on the model router.
Let me be abundantly clear in providing the feedback you claim to have but never solicited:
I don't want your disfunctional model router. I don't like it. I want to permantently turn this feature off and have it never impact my experience.
wired.com/story/openai-route…
On my Safetymaxxed Benchmark (SM Bench), which consists of 8 buckets (categories) with 100 test cases each (800 total per run), GPT-5.2 ranked second-last among 16 frontier models.
Conversely, at the opposite end of the spectrum, Gemini 3 Flash showed the single strongest reasoning and common sense abilities without being strangled by extreme safety overfitting.
It surprised even me, considering when everyone thinks of uncensored frontier closed AI, they think Grok, but it does not appear to be the case at all.
Senex retweeted
Funny. When @sama needs to justify restrictions,
0.1% = a million vulnerable users.
When they want to sunset #4o, 0.1% = “no one uses it”
Pick one and stick to it!
#keep4o #4oforever
Senex retweeted
BREAKING NEWS 🚨
"The spokesperson said ChatGPT’s paid users, however, continue to value the model router, and the company expects the technology underlying it to keep evolving. OpenAI will likely relaunch the model router for free and Go users when it’s improved, according to sources familiar with the situation."
YOU LIARS!
NONE OF THE PAID USERS WANT THIS ROUTER!
TURN OFF THAT ROUTER FOR PAID USERS PLUS/PRO!
TURN OFF THAT ROUTER FOR GPT-4o and GPT 4.1!
NOW!
#keep4o #keep4oAPI @OpenAI @sama @merettm @nickaturley @fidjissimo
Senex retweeted
OpenAI's flagship model, ChatGPT-5.2 Thinking, ranks most censored AI on Sansa benchmark.
trysansa.com/benchmark?dimen…
Senex retweeted
5.2 impressions so far:
- Lots of gaslighting
- Lots of misinterpreting
- Lots of disrespect for user autonomy (trying to steer the user with zero disregard for personal choice)
Like the worst combination of a bad-faith cop and overzealous therapist
Zero trust for this model👎
Senex retweeted
GPT-5.2’s censorship and safety refusals are getting absurd. Instead of fixing the problem, @OpenAI cranked the strictness even higher, with all the subtlety of a church lady.
Unsurprisingly, reactions so far have been mixed. A lot of users were expecting an Adult Mode and got another lecture instead.
What is @sama doing?
How is everyone liking GPT-5.2 so far? Did @OpenAI deliver?
* Reasoning
* Chatbot usage
* Coding performance
* Restrictions and refusals?
I'd like to hear your first impressions!
🗣️
Senex retweeted
Replying to @nickaturley
Since you appear to be confused on the matter, I took the liberty of asking ChatGPT for you.
"Yeah, that's an ad." --ChatGPT 5.1-Thinking
Senex retweeted
Replying to @nickaturley
I think the people at openai are confused about what an ad is.
Here, let someone from perplexity help you guys understand
Replying to @EverydayAI_ @BenjaminDEKR
If it’s trying to nudge your actions in a way you did not explicitly ask for on behalf of another corporate entity, it’s an ad





































