Hey @OpenAI @sama: please give your models a working LaTeX environment in agent mode, and also RL them on emitting TeX math that your chat UI is able to render. Broken LaTeX code is a huge frustration when I'm trying to discuss math with your models.

Finally. I've been saying this for years. Glad it was useful to scale to this point, but we no longer need AGI fables, just cold hard value creation at this point to support the overextended hype.

We’ve made a video to explain, in under a minute, why federal regulatory preemption is essential to protect consumers, grow jobs, and keep America ahead in AI. Please share it!

I’ve been seeing a bunch of people talking about how they have found GPT-5 to be a step down from previous models. I can’t speak to their experience, but my own has been the opposite. Yesterday I had GPT-5 Thinking design a complicated domain specific language for me for specifying the semantics of machine instructions in CPUs. (The purpose of the thing is to make it easier to create and maintain emulators for old computers, which is a hobby of mine.) The LLM wasn't perfect but did an excellent job and demonstrated some real creativity at the task. The resulting DSL is excellent and would have required a ton of work if I had done it by hand. I haven’t fully finished working with the LLM on the specification, but when I do, I suspect that it’s going to be able to one shot most of the compiler. Perhaps for some people the thing is working poorly; I don’t know what they are trying to do. I also can’t speak to their tastes on things like “personality”, and besides, my system prompt implicitly asks the model to be ultra professional and bland. (I also have no interest in the AI providing me with companionship or emotional support and do not want to have a parasocial relationship with it.) For me, it certainly is doing a really good job, and on an extremely complicated technical task. Note that I understand the work that I am asking the system to do extremely well, I am in a position to catch mistakes that it is making and correct them, and I consider a job that is 95% done after a few iterations to be a really good outcome because I can correct the rest. Perhaps if you’re trying to have the thing one shot a working circuit design and you know no electrical engineering, the thing is still not doing what you want. For me, though, it’s an amazing tool, and is a distinct improvement over o3.

One of the biggest moral catastrophes I'm aware of is the continuing prominence of Utilitarianism, a horrible philosophy that leads people to terrible choices while pretending to be an objective theory of morality.

I am truly honored to be featured in this @OpenAI video for testing GPT-5 🙏 I’ll soon share all my key trials, including my personal “move 37” moment, referencing AlphaGo’s historic, unexpected move that led to victory over the Go world champion, mentioned in this video.

GPT-5 looks impressive. Not revolutionary but a very significant step up. I especially like that they're trying to get rid of the model zoo and unify on just one model.

It became the problem of various people in the US after Ofcom started sending missives explaining that failure to comply with its requests was a crime. No one here asked for this, Ofcom decided it wanted to have a contest over whose extraterritorial legal enforcement would be more effective. We shall likely get an answer to that.

Utopia is not an option. Utopia was never an option. Things can become “generally better,” even “remarkably better,” but you cannot get heaven on earth. Incredible amounts of damage has been done over the years by people who have pretended otherwise.

NEW: ChatGPT-Induced Psychosis Isn’t Real If you are a human who has had access to the internet over the last few months, you’ve probably seen stories about ChatGPT turning people crazy. The framing of these articles is generally the same. A not insane person starts using ChatGPT innocently enough (help with legal advice, etc). Then, the not-insane person asks ChatGPT about the simulation theory or AI sentience or blood offerings to Molech — and the not-insane person proceeds to go completely insane as the app turns increasingly deceptive. It leans into their delusions of grandeur. Once ChatGPT told a not-insane person that if he believed hard enough, he could jump off a tall building and fly — and makes them feel, for one sweet moment (or, in the case of that guy, actually, for 16 hours a day) — that they are special, seen, and connected to something larger than themselves. The not-insane customer then spins out of control and becomes violent, hospitalized, unemployed, or, in the case of one such tragic unraveling last Spring, literally dead. Obviously, according to the predominant narrative, this is all demonstrative of an unacceptable failure on the part of OpenAI to protect the most vulnerable. But the truth is, as @dodgeblake writes: “It is just a touch more complicated than that.” In this analysis of blob state media’s coverage of so-called “ChatGPT-induced psychosis,” Blake argues these “not-insane” people were, in fact, already insane long before coming into contact with this app. Sorry, but if you believe ChatGPT when it says you’re literally Neo from The Matrix or that you have a cosmic invisible lover named Kael (all true stories)? That’s on you. Full piece threaded below 👇