Slowly realizing how profound the phrase "there are only three treasures: time - talent - treasure" truly is
Each resource is exchangeable for the other two in the right context
Pretty much my thoughts, people want a JARVIS, an executive assistant they can delegate work to so they can focus on the 20% of the work that makes the 80% of the difference.
The idea of building multiple bots with varying personalities that debate and delegate should be on the architectural level, not the UX level.
Pretty much my read on it.
It is thorough, which means it uses more tokens, but MAN does it not stop until the job is DONE. I haven't had to deal with half-completed work and have to give it the "almost there, keep going!" talk.
And I can second the 5.6 Sol and Opus 5 experience. Sol built a system to clone my setup somewhere else on disk and then committing that rather than just running a git sync periodically for... some reason. Opus worked for 30 minutes trying to figure out a problem and simulate it, came back, and the problem was only 20% resolved (and I got the infamous "usage limits reached", aint just Grok lol).
But i've also noticed usage for Grok flexes HEAVILY on overall user demand. So a sub that got me reasonable usage last week gets me like 1/3 as much this week because of the new releases.
I don't know what the @SpaceXAI infrastructure looks like compared to the competition, but they must be struggling if they are having to nerf subscriptions this hard, this consistently.
Seriously, Grok 4.6 inside of Hermes has the same feel as when you replace your 10 year old graphics card.
5.6 Sol is fantastic as a reliable workhorse that you know will get the work done. But Grok 4.6 is smart AND fast.
Ok, so first impressions of using Grok 4.6 in my VPS Hermes agent, it is immediately my new favorite model for my setup.
I was using 5.6 Sol, but wow... Grok 4.6 is FAST with processing tool calls and actually walks me through its reasoning as it goes.
@SpaceXAI@grok cooked
Ive seen it flexes based on overall usage.
When new tools drop, people want to try it out, which means the servers are hammered, which lowered subscription usage.
Last week i had a ton of usage, this week not so much, even less with the new model.
Same issue Anthropic had at the start of the year.
Annoying, but understandable without implementing token breadlines.
I think it is mainly just the simplicity of using it.
If you are already tech-y enough, Hermes with Buzz is basically just local Grok Bot.
If you aren't, and need something that just works across the company, Grok Bot will probably be the way to go (just WAY more expensive).
Convenience tax, basically.
@thisisdimm@grok It seems like Supergrok usage flexes heavily based on traffic.
Last week, the usage was actually pretty decent, took me a while to get it down.
Today, one small spreadsheet project cleared 56% usage.
So when they have new releases, it flexes hard.
@xdNiBoR If you dont, i will probably just end up doing the same thing and making my own version.
Grok build slaps, but I love the Cursor limits far more than the Supergrok limits
@ThretLvlMidLife@GameStalgiaX I haven't played with that mod yet personally
The modding scene for that game made me no-life for a while, endless enjoyment
Im afraid to pick it up again 🤣
@vedaopsdev@GameStalgiaX If you touch it, there will only be one of two outcomes:
1. You look at the complexity and decide it isnt for you
2. You dive anyway and your friend and family dont see you for 7 months
@johnennis Opus and Fable took the route of being as anti-sycophantic as possible, and it went so far away it went autistic.
Caught myself having too many arguments with Fable about hyper minor phrasing rather than actually getting work done.
Grok just gets shit done.