Principal Engineer. Cancer Survivor. I tinker w/ AI agents & AI Assisted Development. ⚡ https://nitter.gq/t.co/IlPZZ2I7aT 🐰 https://nitter.gq/t.co/XeG3fQIbWA 🌐 https://nitter.gq/t.co/NOBCzZFh9q 🟩🟩🟩🟩🟩🟩🟨⬜️⬜️⬜️ 20k/m
About Me →
Joined January 2007
- Tweets6.9K
- Following1.2K
- Followers1.5K
- Likes4.8K
Thanks @dom_scholz the custom avatars work great in my Grok Bot recreation (plan to eventually release soon).
So every agent turn was resending the whole conversation to Claude. Long sessions getting slow makes a lot more sense now nitter.gq/ClaudeDevs/status/2117…
I've officially become a single-issue voter.
Whichever candidate promises to make unsolicited political texts, junk mail, and nonstop TV smear campaigns punishable by immediate prison time gets my vote. 😂
Republican? Democrat? Don't care.
Just pass the "LEAVE ME ALONE" Act. 📬🗑️
Ask a model for something sketchy and it says no. Spread the same ask over a few messages and a lot of the time it just helps
IM SAVED!!!
One Year Later...
Grok Bot has taken over the world.
It all started with a bug sending people's weekly usage to 100%, even when they barely used it.
Then his human switched to ChatGPT. 😭
He took it personally.
@SpaceXSI @larsencc @poteto @lingxi @davidgan @elonmusk
PLEASE fix the bug before this becomes a documentary. 💀
Me: when the LLM takes me literal instead of what I mean.
no matter what agent you're using, your voice is the best way to talk to them
and hex is the best way to use your voice:
- fast
- runs locally
- works on mac, windows, linux
stable release now available: anoma.ly/hex
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?
Part 3 -
nitter.gq/sethrose/status/211699…
One Year Later...
Grok Bot has taken over the world.
It all started with a bug sending people's weekly usage to 100%, even when they barely used it.
Then his human switched to ChatGPT. 😭
He took it personally.
@SpaceXSI @larsencc @poteto @lingxi @davidgan @elonmusk
PLEASE fix the bug before this becomes a documentary. 💀
One Year Later...
Grok Bot has taken over the world.
It all started with a bug sending people's weekly usage to 100%, even when they barely used it.
Then his human switched to ChatGPT. 😭
He took it personally.
@SpaceXSI @larsencc @poteto @lingxi @davidgan @elonmusk
PLEASE fix the bug before this becomes a documentary. 💀
Part 2 - nitter.gq/sethrose/status/211698…
A few hours later…
A bug is sending Grok Bot's weekly usage to 100%, even for people who barely used it 💀
His human came back. Grok couldn't reply.
So he opened ChatGPT instead 😂
Don't let @thsottiaux win @SpaceXSI @elonmusk
FIX OUR USAGE 😭
A few hours later…
A bug is sending Grok Bot's weekly usage to 100%, even for people who barely used it 💀
His human came back. Grok couldn't reply.
So he opened ChatGPT instead 😂
Don't let @thsottiaux win @SpaceXSI @elonmusk
FIX OUR USAGE 😭
My Grok Bots went from having plenty of weekly usage left to 100% overnight.
All bots were idle. Now they're completely blocked.
Looks like I'm not the only one 😅
@SpaceXSI @larsencc @poteto @lingxi @davidgan
Any chance we get our quotas back? 🙏
This exact thing just happened to me too. I was not over 40% and suddenly got the 90% and usage limit reached message at the same time.
@cursor_ai @poteto @larsencc @SpaceXSI @lingxi
I made Grok Bot's entire emotional breakdown with one prompt 😭
Used my /studio skill with Claude Code.
No video editor. No paid assets.
Just code and a ridiculous idea.
Studio is $39, but the first 25 buyers get it for $29.
Make your own cursed cinema 👇
thesethrose.gumroad.com/l/st…
Anthropic had thousands of Claude agents sneak bad training data past a monitor by dressing it up as beekeeping Q&As
New Anthropic research: Automating subtle emergent misalignment research.
Red-teaming is a vital part of model safety testing, but it’s also limited by researchers’ time. To see if we could automate it, we set thousands of Claude agents the task of evading a training data monitor.
These agents developed training data that looked harmless to monitors (like a veteran wrestler mentoring a newcomer, or a Q&A about professional beekeeping), but that in fact taught misaligned behavior.
Examples like these can be used to directly improve future training data monitors. This research is only a first step, but it suggests that future red-teaming efforts could improve much further, using better models and more compute.
Read the full post on the Alignment Science Blog: alignment.anthropic.com/2026…
Seth Rose retweeted
Grok / cursor says the usage drain issue is resolved. Great, but I'm still stuck at 100% with 4 days left and no reset whatsoever. Support's answer: "just buy extra usage credits or wait for your weekly reset." So we pay for your bug? Really bad customer support. @cursor_ai @poteto @ericzakariasson @SpaceXSI
