Wisdom that's been lost in many modern quarters. I'll add that, if you need to take off the coat, you have several options to keep the pocket gun in your immediate control:
1. Hang the coat on the back of your chair (assuming no kids around).
2. Transfer the gun to an empty pants pocket. Get a good pocket holster.
3. Transfer the gun to your waistline. For the waistline, @HamreForge grips with their "Hamre hook" help immensely.
The vendors & partners that made it possible.
@sendcutsend@ARK_Electr0nics@HamreForge
What do they all have in common?
They're fast, they work with anyone big or small, and they're American ๐บ๐ฒ
Don't worry, even if you have a SuperGrok Heavy account, reach out to https://t.co/2r50LkWBCd and send an email to [email protected] or [email protected], or submit a ticket to the sales website, or submit a request through the support portal, you will still not get an answer from anyone. @grok will just make stuff up. We've been dealing with this for about 2 weeks now.
.@elonmusk is intentionally or unintentionally moving the goal posts with @spacexai and @grok. As a small business, it's frustrating, and starting to impact how we work. The other difficulty is that there isn't anyone at https://t.co/ukJp4oYXQl who is responsive to any inquiry about usage limits, and models.
Grok has been getting increasingly more expensive, and additionally, ineffective for us as a business to complete tasks, as we previously were. I have an example.
Back in February, I used @SuperGrok to create a spam filter and email categorizer for me that ran against a gmail account. This gmail account isn't an "active" gmail account, I use it to sign up for things or for web purchases and receipts. It's a throwaway. It has pretty consistent spam, receipts, and usage patterns. With this api call, I haven't made any changes to it since February.
It has been running the same app, the same API calls, the API has been absolutely untouched, and usage has been the same.
Attached are 3 screenshots.
One of them shows the graph over time since the app was installed and goes through until today. Exact dates are February 24, to September 25th. In early March, there was a usage spike, due to some change with X / SuperGrok / Grok. Then sometime around the middle of May, 2026, spend and token usage went through the roof. Keep in mind, this is an extremely consistent environment, with static code in the backend making the same API calls over time.
Another screenshot shows a one week snapshot in March. Dates between March 4 and March 11, 2026. A total of 209 requests were made, using 201,974 tokens, with a cost of $0.05.
The third screenshot shows another week snapshot in time, but between July 19, and July 26, 2026. A total of 192 requests were made, using 191,481 tokens, with a cost of $0.25.
For my usecase, the task is the same. The results are the same. The volume is the same. The code base on my end is the same. The API calls are the same.
But the cost is 5 times where this app started.
Now, we've been a commercial account, with multiple Supergrok Licenses, and a single Supergrok Heavy license. Our invest has been substantial for our small business. Starting 2-3 weeks ago, our grok chat is unusable. We are getting rate limited within a week. We don't do image/video generation. We're a small manufacturing company that uses an LLM the way it ought to be used.
We need transparency on why/how these changes are being made to our accounts. When we purchased an annual subscription, we purchased a product that worked for us when we bought it. We allocated funds to get this product for a year. But now, we're not getting what we paid for. We're getting limited in under a week. We get told to wait for a weekly reset for the token limit to be lifted. But then when we waited for a reset yesterday, we try a prompt after the reset, only to get a response back that we have to wait for our reset on October 1. So a full week where we can't use our Supergrok and Supergok Heavy licenses for business. Unless we buy even more tokens, that just get burned up again in no time.
It never used to be this way. We'd have chats that were thousands of prompts long. We can't do that today.
This is unacceptable. I don't know what changed, but things have, and it's hurting small businesses.
Maybe it's intentional, maybe it's a bug, but whatever it is, it's not good.
We're a small manufacturing company, but with 4 annual licenses, 3 SuperGrok and 1 Heavy. We only use chat on https://t.co/0w8APqMAkN and in app. Our business team hits weekly limits by just sending a few hundred inquiries, via chat. Not complex asks, not deep coding. No image or video generation or voice. Simple chat, using millions of tokens, chewing through hundreds of thousands of tokens per message.
It's ridiculous.
@Judge_Me_Man@lydiahallie No, I have heavy, still hit these crazy limits. I never used to, but started hitting limits with September updates x made to @grok
@LMantalos@ScottyBeamIO We don't even use Grokbot, just https://t.co/cTyAMEcH4N and chat. Hitting limits for the first time with tier 3 plan. They made some changes, and nerfed usability.
Hey @elonmusk, changes to grok have slowed our small manufacturing business to a crawl regarding innovation. We only use https://t.co/z2FmZvXNgh and chat. @grok usage limits have kneecapped our work. Grok recommends switching to @chatgpt to return to our normal usage patterns.
@QoraAPI@JasonBud What we're seeing is essentially this-https://t.co/VEYzRHCpHf
But we're not doing anything that should be token heavy. Just menial tasks and getting limited from the usage
Been using Grok 4.7 a bunch today and I really really like it. It's snappy, follows instructions well, and the "fast mode" is very fast
But price is a huge issue...
1. I don't have good data for this yet, but it's pretty clearly not all that efficient (esp next to something like Astra)
2. The pricing really isn't the same as Grok 4.6. Over 200K tokens it doubles and with fast mode it increases even more
(attached screenshot is from grok build analyzing it's own usage from today)
3. The usage limits at least in grok build are BAD. I'm on the $300/mo. supergrok heavy. I've used ~$40 and it's burned ~8% of my weekly limit. This is nothing compared to what u get from OpenAI or even Anthropic...
I haven't been pushing it super hard in there yet b/c I'm on crappy airplane wifi, but I'm really worried I'm gonna rip through the limits at a brutal rate. One of the best parts of previous grok models was you basically got unlimited usage in Cursor/Grok Build on the big plan, that seems to be done.
Which imo hurts this model way more than it hurts Fable b/c Fable (or Astra) is a true "premium token". It's so good I'll deal with the harder limits, if I'm getting limited on Grok I'll just go use Sol or Opus (well not opus but u get the idea) since this "tier" of model has tons of competition
It sucks b/c I like the model a lot, and I really like the surfaces it's used in. Grok Build is the best CLI, Cursor's cloud + project system's really nice, and Grok Bot is a huge part of my main workflow at this point.
They really need some of that OpenAI efficiency magic to stay competitive at this tier, or start making much bigger more powerful models
We have primarily 2 users, me and 1 other, while there are 2 licenses allocated, used little to zero. Audit logs validate this. Prior to September 11,we never saw any spikes. Starting Sept 11, @grok is chewing through tokens. 768 requests, just in https://t.co/QRywcgeLOO 143 million tokens. We haven't done anything different, no api, just Grok chat. We're a small manufacturing shop