Flowers ☾ ❂

There will be new, original, unified GPT-5 models. People who say they know that it will just be switching from 4o to o3 are just reproducing each others yapping. It might be the case that GPT-5 knows when to think and when not, but they are completely new models. PERIOD sigh

Dont believe anons hyping up that a model won the fucking IMO gold medal, but please do believe me that my anonymous researcher friends arent satisfied with it 👉👈 Same guy who said Grok 4 is AGI btw (he deleted the tweet shortly after Grok 4 release lmao)

Is o3 alpha the open source model, GPT-5, o4 or something else?

Agents computer btw

Agent drawing a robot

Agent creating a doodle jump clone. Since it can see and test everything itself, it can fix the code iteratively until its finished. Also, the game elemtents look really good, better than with o3-pro, Gemini, Opus or Grok

Who said agent mode doesnt come to EU ;)

The only meaningful way to benchmark AI models going forward is through games. Forget knowledge based tasks, these models already know everything. Games are the real test. They demand everything: rule following, working memory, abstraction, planning, lane switching, even a form…

ughhh no matter what I try to do to create a new benchmark, o3 always immediately makes it meaningless

I also said no Opus 3.5 and no GPT-4.5. Both were right at the time even tho Anthropics OWN website mentioned Opus 3.5 GPT-4.5 was meant to be called GPT-5 at the time (as per WSJ) Never bet against Flowers.

Will they call GPT-6 GPT-6 or will it be just an update for GPT-5, or will it have a completely new name. How will GPT-6 equivalent be called?

If you didn't think your company was the closest to AGI, why would you turn down a billion dollars

So is anyone actually using cluely or was that all a week of hype for literally nothing?

Don’t believe everything you see on Twitter, kids. Most claims on any topic are just engagement bait. Always work on sharpening your critical thinking and cogsec and don’t let random anons rile you up like a mob with pitchforks. I still need to get better at that myself.

I have a strong sense that something huge is going to happen this year. Something extreme. I don’t know if it’ll be good or bad. I don’t think it’ll be about AI, maybe it will, but something’s coming with a magnitude beyond what we’ve seen in recent years or in the near future.

I asked ChatGPT to invent a new Latin phrase that sounds ancient and wise and describes itself or LLMs in general. It came up with Verba sine mente, scientia sine fine. Words without mind, knowledge without end.

AGI will be able to play any game and do any digital task and there will still be some people saying that this isnt really impressive because this and that

Is the experimental model that won the gold medal at the international math olympiad AGI?

OpenAI’s lab models aren’t far ahead of public ones! Except ofc those staying under wraps for another six months (which is about 5 years in non-AI time)

You can make ChatGPT reply in red btw. Write answers using LaTeX math mode with \textcolor{red}{...} inside $$...$$.