Claude Opus 5.5 Just Did Something That Looks Like AGI

Checkout todays Sponsor – https://chatllm.abacus.ai/min

And it still cant combine 3 excel sheets with formatting. 'AGI' STFU

Opus 5 was infuriating. It wasted so much of my time, that I went back to opus 4.8. I had Opus 5.5 work on my startup code yesterday, and man… it COOKED! The strategies of "ask me as many questions as you need", in plan mode, and pushing it to research the systems near what I wanted to build upon (I gave it the context of what those systems are)… It built, it tested, and it took my AI layer (which has text to text, text to image, text to video, and image to video models) to the next level. Now I have multi provider support, including fallover, with site admin ui configuration… It just worked with little guidance. Given there's a ton of work I've done to create guiderails for what I expect in terms of code quality, and worked to build and maintain a quality codebase. But compared to opus 5, it's a completely different world. Most of this code btw was built with opus 3.8 and then opus 4, before they nerfed opus with 4.1 a few weeks later. This is a startup with over 200k of ARR and just me and a CEO. It's a wild time.

For what I'm doing, I'd want to split the work into tasks based on use case with a interoperability contract. Leave the coding and design to Opus 5.5 and the engineering and debugging largely to Astra.

1:30 yeah it's almost like there's something a little fishy going on with these benchmarks. Fable came out a few months ago, there's no way their opus model is gapping it by this much. They're benchmaxxing.

0:56 correction: OpenAI says the Navier–Stokes result was produced by a new internal model “significantly more capable than GPT-6 Astra.”

ABBA-cus

Opus 5.5 is nice to use, fast, drinks tokens at a more sustainable rate, does what I ask. The internal browser use is fast as well. I think the only thing keeping it behind OpenAI now is the voice support is not nearly as good. They need to fix that. Otherwise, it's pretty much the best AI money can buy right now

Looks like ai still hasn't found a cure for your rampant adhd 😂

AGI 😂😂😂, do you even understand what it is? If Opus is AGI, the my toaster is ASI.

Lost all respect for you calling it agi

Perhaps not a "new breakthrough" in Opus 5.5, perhaps something like following GPT 6 Astra's lead in implementing latent space recurrent depth.

calling one impressive model run agi is benchmark theater; repeatability and autonomous goal-setting matter more than a viral demo

F F S…

Every new model is a tremendous/YUGE jump. One giant hype machine.

This comment might not age well but is it correct to assume that AI generation of video games will carry on exponentially for the next year or so?
Maybe there is a limit in terms of what it can do unassisted plus.. if its IQ is based on watching games stream on Twitch, etc., does improvement assume that video streams will be easy to access in future?

Bro can't read abacus

astra is over rated. opus 5.5 is clearly better