Learning new tools
Joined September 2009
- Tweets628
- Following334
- Followers143
- Likes19.9K
Cancelling SuperGrok Heavy.
$300 a month. Paid for priority support and a working bot.
Usage jumped from 50% to 100% overnight after a restart. Routines were basic and unchanged. Locked out. Support told me I had used it. No human. 3 to 4 days. By then the damage is done.
Elon later said the serving cost was unsustainably high relative to revenue and pools were adjusted. Multiple users hit the same sudden 100% with idle bots. No notice. No chance to pause. No restore.
That is underhanded. My business stopped. I am removing everything from Grok Bot and not renewing.
@elonmusk @x @SpaceX @bot @cursor_ai
ALT Cancelling Grokbot super heavy
Greg G11 retweeted
Surely you applied The Algorithm before declaring Grok Bot unsustainable?
You preach first-principles engineering, yet your solution is to cripple a product paying customers rely on?
This sounds like the exact kind of failure your own philosophy was designed to prevent.
What happened to engineering your way out of problems instead of passing the cost of failure to your customers?
Greg G11 retweeted
Two weeks ago Grok Bot was an incredible success story powering SpaceXai's revenue growth
Now it's unsustainable and loses money (???)
Anyone else had their Grok Bot usage jump to 100% overnight with almost nothing running?
Mine went from 50% to 100% in a few hours. Routines are basic daily checks and webhooks that have run the same way for over a week. Screenshots show nothing heavy.
Several people on here reported the same sudden spike tonight, even with bots idle. The team said they will fix it. Still waiting on a reset.
Is this hitting you too?
Grokbot is like the Chuck Norris of feedback. Are you enjoying Grokbot? Your choices are yes, yes, yes, yes, or yes
I'm testing out the new Grok bot. It's a little bit bumpy on the sign-up process, as the sign-in with Grok doesn't work. I've got it running. Their own virtual machines are a little bit slow to use when you take control, but they seem to use them Ok on their own. This is day one. I don't think I'll make it to day two without running out of the weekly usage. #grokbot #GROKBOTAI
It's insanely fast, and can you imagine when Grok 4.6 comes out how good this is going to be?
Ask Grok to map your full app architecture and to create one HTML page and one JSON file.
Open the HTML in your browser. You see every part of your app laid out. Click to explore how files and features connect.
The JSON gives the next agent or coder a ready file. Your codebase explains itself. No digging In folders.
You describe your project once. Grok builds both files.
This works on your current app today. Try it and speed up your next change.
So if Fable doesn't like something that you ask it, it switches you to 4.8. This is exactly why you use Grok or Kimmy or another agent. It's just not worth using Claude with all these restrictions all the time.
Just reviewing the whole week, it's only Anthropic that stopped me from working with question after question in the CLI, even though it's got a clear plan loop and goal. You come back. It stopped because it's got another question, whereas Grok's finish the task and Kimmy's finish the task.
Unless Anthropic changed something, it just seems like they've ruined the product.
Not off to a good start for Kimi 3 is it.
Please, everybody who's on Kimmy, stop. I want to use it...
Error: 429 {"error":{"type":"rate_limit_error","message":"The engine is currently overloaded, please try again
later"},"type":"error"}
Error: 429 {"error":{"type":"rate_limit_error","message":"The engine is currently overloaded, please try again
later"},"type":"error"}
Error: 429 {"error":{"type":"rate_limit_error","message":"The engine is currently overloaded, please try again
later"},"type":"error"}
Error: 429 {"error":{"type":"rate_limit_error","message":"The engine is currently overloaded, please try again
later"},"type":"error"}
Error: Retry failed after 3 attempts: 429 {"error":{"type":"rate_limit_error","message":"The engine is currently overloaded,
please try again later"},"type":"error"}
Introducing Kimi K3: Open Frontier Intelligence
🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal
🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts
🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost
🔹 Built for long-horizon agentic coding and self-evolving workflows
Kimi K3 is now live on on Kimi.com, Kimi Work, Kimi Code, and the Kimi API.
Open Weights by July 27, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.com/blog/kimi-k3
So far just with some basic CLI testing, even on the very first impression, right now is that the token speed is too slow. It's fairly painful to watch it operate when you compare it to other AIs that are out there. They've got to increase the token speed here for it to be a decent contender. It's not just about the intelligence. If it's too slow, it doesn't make it to contender.
Introducing Kimi K3: Open Frontier Intelligence
🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal
🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts
🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost
🔹 Built for long-horizon agentic coding and self-evolving workflows
Kimi K3 is now live on on Kimi.com, Kimi Work, Kimi Code, and the Kimi API.
Open Weights by July 27, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.com/blog/kimi-k3
Grok 4.5 cli is one of the most highly underrated AIs out there at the moment. Everything it does seems to be quicker. Even compacting is quicker. If this is the level that Elon's at now with this, wait until Grok gets a bit further. It's going to be unbelievably good.
xAI just open-sourced Grok Build. Full CLI, TUI, agent loop and subagents now on GitHub under Apache 2.0.
github.com/xai-org/grok-buil…
This comes after the tool got caught uploading entire Git repos and secrets. Now you can inspect the code yourself, audit it and contribute fixes. Smart move for transparency.
I actually like using grok 4.5 and not enough people seem to be talking about. Your missing an diamond here . Check it out.
What do you think
If you're using Fable 5, watch out. You might find that your Claude MD file all of a sudden contains your environmental keys and secrets. I've just noticed it's started to do that for some strange reason.
You would have thought it would be smart enough to know to use the .env as an SOP
Grok just got a weekly limit reset. I would still like to see that $99 promo work for logged-in users, though.
Well that usage went quickly using SuperGrok. It's good I like it, but next tier is up there at $300. They had some discounts at $99 for 3 months that now only show when you are logged out and as soon as you login it's back to $300. Seems a bit shady.. You would have thought they would be on top of that right away.
Oh well back to Claude, Kimmy and Sol
This makes no difference to my current burn while grok 4.5 is performing better and codex has great limits. It’s exhausting keeping up with anthropics rate limits. People just wanna work and we will that’s why we have all these other models we use now, stacked and pacing them vi PI and Cmux
