Why I paid Grok $200 after its limits pissed me off

I tried Dots, Hermes, Muse, Insight and Duet. However they started, I kept ending up with a chief of staff checking my email and calendar. I gave Dots hundreds of attempts and reset it four times. Grok Bots became teammates. I put them to work on startup funding, marketing and content, AEO, measurement and builds. When the allowance ran out, work stopped. That’s why I went from $50 to $200.

I paid Grok Bots to do work I could hand off

My Grok Bot assignments crossed several kinds of work: startup funding research, marketing and content, AEO, measurement and builds. It felt more like a team I could put to work across a brief than another assistant sorting my day. I could judge the result by whether I could pass it to someone else.

On November 11, I put Grok Bot on the weekly SEO and AEO report for Kitsune's site. It compared 31 exact searches with the previous report and found that none of the top-five web-search results for 30 non-brand queries showed Kitsune or OdinVision. Our own name returned five of our pages in the top five. The report called out one especially useful gap: the new AI visibility check was absent from the top five for a Pakistan query asking which brands ChatGPT recommends. That gave us a concrete buyer question to work on.

That report used web search and a synthesized answer as a proxy. I checked the result and used the buyer question to shape the next piece of marketing work.

The company says the work is expensive to serve

The allowance still matters because the company has to pay to run this work. On November 1, Elon Musk wrote that “the cost of serving Grok Bot was unsustainably high relative to revenue,” as users complained about hitting their limits. TechRadar reported his reply and linked the X post. The next day, xAI announced Grok Bot 1.0 and said it had made the Bots more efficient, using fewer resources to complete tasks. That is the company's account of the release.

On October 12, xAI announced The X Subscription, combining Grok, Grok Bot, Cursor and X. Its $200 Ultra plan lists fifty times the usage of the $8 Plus plan. The company's pricing documentation says Bot usage comes from a shared weekly pool. xAI's announcement sets out the plans, and its pricing documentation explains the allowance. And for someone like me, its important to be able to utilize my biggest movers to the most of their ability. At my under $50 subscription, that was just not cutting it anymore.

The people selling agents can show plan price, seats and usage. A buyer should also track which deliverables came back, what still needed a person, and where the agent stopped. Those are the units that let a team compare the invoice with the work.

Stop selling me more "cheap of staff" agents

As a builder and someone who loves to buy the year outright, I want more and more of my agents to be available to me where I am. Like real people. On the phone. And for them to be able to do a multitude of tasks independently or in coordination with real working harnesses. And that works wonderfully with Grok Bots and Cursor. On the other hand, despite even making shared ledgers, notions and even local SQLs, Dots would forget how to do something within a couple of hours of having set it up.

If you are new to this, I would test one recurring job from the start of the week to the moment it is done. Give it the source material, boundaries and expected format.

I paid more because the work stopped when the allowance did. The price is the easy part to compare. The harder question is what I can hand to someone else to take it to the finishing line.