it is superior almost

This is the other side of AI too, they are used to act like that even before AI explosion But now it is more evident then before!

Have you heard about how to instruct an llm on writing clean code and architecture. Why it should be am issue in five year? Now llms are capable more than a senior SWE, imagine in 5 years, maintaining a codebase would not a problem, as it isn't now

Are you a Vannaci's fellow?

Another agent aligned in reviewing

You can always cleanup the ones not needed anymore uh

Nope, I do have prefilled orchestration graph system with skills that runs it instead of letting the orchestrator deciding what to choose as classifier or reviwer, I don't like the idea to let the model deciding it, but maybe I'll develop a variant for my graph just to try it out

Se I bond sono inflation linked saliranno Ma comunque chi cazzo compra i BTP con sta situazione? Bisogna essere dei pazzi al di là di tutto

You know that you could have the second one inside the first one? No cheating.

You have to work on your harness man. For sure you have something wrong going on

Are you talking about the weekly limit or the 5 hour window limit? Because if it is the first then you're right, in the second case no because with high effort it drains quite fast tbh

It depends, for example I use opsu 5.5 and codex, this way I don't run out quota never, I use both simultaneously on multiple projects with multiple agents, so good!

If you wanna try it on opencode and using pi I've built a bridge for it here, check it out! github.com/Gybra/pi-opencode…

It depends what the task is and how big is the codebase, I did some improvements to an open source repo of mine and it actually did very well. So i don't get the point, you can not pretend it to act on big codebase like opus 5.5 for sure

So you're one of those interested in the last attack?

In quel posto di lavoro ci entra un ragazzo.. Si ma che tipo di lavoro parliamo? Stem o da parassita nella pubblica amministrazione? Ste cose non le valutano.. Ovviamente cazzo ce ne frega della cultura.

2K for Mac Studio compared to 128Gb RAM DGX Spark is not that fair OBVS DGX Spark wins this comparison

For sure, i've found out that most ppl out there dunno much about all this revolution and how eg: agentic workflow works. I'm still teaching to my colleagues how to create a graph for an optimized daily agentic workflow

We could start right away from the ones we have available rn as a baseline models to improve'em and build on top our owns

Yes that's insane and i'm already building a bridge for that for pi coding agent 😍😍😍😍