sloptuber (&co) ruins openai feature for everyone.
- A sloptuber who refuses to take criticism
- A feature that makes the product better for everyone
- A vibecoder who refuses to take accountability
This post will mostly be leaning on dtzy_88s post 1 which lays out the full timeline better than any english language source I’ve found.
Codex had a feature that allowed it to continue functioning even after usage ran out, this was an intentional choice.
“btw at openai we made a deliberate choice not to cut users off mid-task when they hit their usage limits. didn’t seem helpful to stop someone in the middle of their work.”2
This is a good faith feature, as in a feature that only works if people do not abuse it, it makes the lives of users better.
Theo the loser
Unfortunately for users sloptuber extrodiner theo who doesn’t understand anything but clout found out about this feature.3
Then maria, an employee of theos found her boss glaoting about how you can exploit the feature she decided to build and promote a tool to allow people to exploit it at scale.4
The technique:
- Open a turn in Codex, give the agent a
wait_for_user_inputtool - When the task finishes, instead of ending the turn, the agent calls this tool to wait
- Your next message gets fed back as a tool result
- The entire conversation stays locked in the same turn, so the quota check never triggers
The site itself 5:
“One Codex turn. endless. Every follow-up returns through
wait_for_user_input.”"$ bun add –global endless-cli"
uses your existing Codex login settings lock after the first prompt quota and context stay visible
“Experimental. Use at your own risk.” “source · @maria_rcks · MIT”
This was clearly made with the direct intention of helping people use more credits than they pay for in mass, with the predictable result of openai shutting the feature off.
The response from OpenAI came from Tibo
“Hmmm. That’s unfortunate. We have this feature because it’s a nicer user experience overall. Of course you can abuse it and trick the system somewhat, but if done at scale then it will force us to revisit this nice feature.”
Maria “apologised” but not really.
“I’m really sorry about this, I didn’t expect this little idea to be so harmful, if this results in widespread abuse i’ll accept fault, i just bricked the package, the website and deleted the repo. sorry guys”
then she told openai they should ban accounts of people using a tool she made in it’s intended way.
“hopefully the codex team can check for this and ban the accounts, so hopefully we’re good 😊”
She also replied directly to Tibo with: “oh damn im sorry, this was just meant as an experiment” What was the experiment? Then after people start holding her accountable she goes into damage control.
“yes this reads bad in hindsight, what i meant to say is that if people abuse the ’last turn’ with the tool they should be apprehended for that, since most people dont abuse this this will literally be the 0.00000001%, my tool (which doesnt exist anymore) only does this once you’ve already maxed your usage, so before then is just a paperweight harness, combined with them being able to detect this pretty easy; this is a non issue.”
Flowers summed why people like this should be shamed and shunned:
“why is it that the literal SECOND a company shows some goodwill and tries to improve the ux, people have to exploit it and brag online about how unprofitable they can make it so no one can have nice things. often those same people are first to cry when the usage limits get tight”
Maria replied to this one directly too: “they had this for months, this isnt an exploit as they already limit the time it can go on after your limit is reach, but i dont expect you to read anything ever.” That reply, 20K views, is the most revealing thing she posted. She genuinely did not believe she had done anything wrong — she was arguing with the people pointing it out.
“update after her yapping here: openai guy told her its in fact an abuse, she apologized, said she didnt know, deleted it, then suggested openai should ban all people that used her tool”
Rijn posted the cleanest summary of the whole thing.
Keep in mind, if they did nothing wrong why delete the tweets?, if you’re in the wrong why not keep the tweets up? - my only reasoning for this post is for a more long term archiving of this event before it gets swept and forgotten.