Reduced Usage on Grok Bot

So, head’s-up, Bot fans. It looks like the usage spike over the weekend wasn’t a glitch at all. It was an improperly-implemented credit reduction.

At the moment it appears the closest we have to an official response on this is Elon Musk saying on X that Bot costs xAI more to run than they get in subscription money. This appears to be based on perceived value, though — that is, on the market prices for the compute involved. So it’s not that they’re losing money, per se; they could be getting more money by allocating the compute elsewhere.

So the going theory right now is that xAI was rebalancing the compute availability for Grok Bot and instead of activating at the next user reset it did so in the middle of various runs, pushing a lot of us to 100%. This matches my observation since I saw a lot of chatter about usage spikes that did not go to 100%, which I began suspecting meant this wasn’t a glitch. It appears I was right, pending official word. Again, all we have right now is circumstantial evidence. If true, xAI really bungled the rollout.

Regardless of the cause, Efficiency Quill is telling me that each run appears to be consuming roughly 25% more usage compared to previous weeks. I’m already at 27% less than 24 hours after my reset.

This is good for another article, one I might write if I have some spare meat-brain compute later; but I think going forward that if xAI wants us to depend on them, they have to be at least as transparent as an electricity provider. Clear data on consumption, tips to reduce compute demand, and pricing based on actual activity. And considering Altman bad-mouthing people who want good home computers, if I had Elon’s ear I would tell him this would be the perfect time to hint at xAI planning some kind of home computing/offline model in the vaguely-near-future.