Musk is third, he's first! OpenAI engineers are back on top of X

  It's the first time in the history of the world.

"People have run out of the net and they're stopping me on the street. I'm going to print a little card and send it along".

This is a post from Tibo, an OpenAI engineer responsible for Codex and ChatGPT products, just posted on X.

A technician in charge of a core product, his exit was seen as a food stamp.

What is real quota demand?

It's the comment section that knows.

A Saif user said that his Codex quota was just restored yesterday, and today it is 37% off. He paid $500 a month for his subscription.

At this rate, the limit for heavy-duty users is a tight suit that can be worn without breathing.

Another user, Nick, used the Pro x20 package, opened the Extra High mode and talked about the five-round allowance.

It's a direct hit to the biggest message length warning. Only five rounds, it's amazing.

Others claim that they only use the high-intensity mode of GPT-6.1 Sol, which is about 10 hours below the limit.

The shortest one came from Oikantik: "Need to reset, please, the context is gone".

In the comments section, people are starting to help Tibo come up with an idea.

Some suggested placing a reset vending machine in OpenAI's headquarters; some suggested simply taking a portable printer, printing out the 2D code, and scanning for a reset card; and others predicted that the card would be brought to the doorstep by Yellow Bull for sale as "a reset of fifty knives, no deal".

It's not enough, but it's not enough.

In September, the launch of the GPT-6 Astra pushed users' appetite to the limit. Demand was so high that OpenAI immediately suspended new subscriptions and upgrades to the $200 Pro 20X class.

Astra burns at a rate about 2.5 times that of Sol.

There is also a way to use it. Once the agent is started, it often works for several hours in a row, and for heavy users, the quantity is the production material.

So the more frequent the repositioning, the more tightly the quantity is pressed.

This explains something even more bizarre.

On October 18, Similarweb released the September X Platform account page traffic share rankings.

The first is @thsottiaux, or Tibo. The second is football transfer specialist Fabrizio Romano. The third is Musk.

You're right, an OpenAI engineer pressed a mask on the homepage visits on X.

In terms of fans, Tibo certainly doesn't compare to Musk, but the number of people running to refresh his page every day to see how many posts he posted today, in August, he was ranked second, and in September, he climbed to the top.

It's all about repositioning.

Tibo, known as Thibault Sottiaux, joined OpenAI in 2024, having previously worked at Google, and was fed up with the bureaucracy of large factories, jumping to a more radical OpenAI.

In 2025, he released the Codex to the public, and by 2026, he had taken over the ChatGPT, Codex, and developer APIs and made them into a super-app.

His X account was registered in August 2025, and by July of this year had gained nearly 250,000 followers. Netizens nicknamed him the "father of cyberspace" and "the god of Codex control".

There are even several civic websites dedicated to tracking his movements.tibogpt.com only answers one question: "Has Tibo been reset?" The answer in the long run is "Probably".

Tiboreset.com has a rearrangement of probabilities and a 28-day challenge to progress. Didreset.com is like an archive, archiving every quota announcement he makes in stages.

Statistics show that from September last year to October this year, he sent a total of 83 quota announcements, of which the total amount was 60 replacements and 12 replacement cards.

But the series has not only laughter, but also a test of trust.

Following DevDay in early October, many heavy users complained that the quota was reduced, the model slowed down, and the comments area began to be questioned.

Tibo's response was to issue a direct military order: for the 28 days beginning on October 5, the team either released an improvement that was relevant to the majority of users each day, or reset it entirely.

The results of the previous 13 days: 5 complete resets, 1 reset card, 7 improvements. The improvements include a 60% reduction in desktop end-disk occupancy, a significant reduction in "model loaded" errors, and a repetition of the context between sub-smart bodies to activate the same amount of activation.

By Day 14, the status displayed ontiboreset.com was "Waiting for Tibo", and the entire network was back on fire.

Repositioning is a pain reliever and improvement is the real problem.

But the reality of the moment is: the stronger the model, the more users can't leave, the tighter the quantities.

The Token is a new era of food stamps, and Tibo is the one who sends food stamps at the window.

28 There is still half a day left for the challenge, and everyone is looking forward to it, except for today. What day is there that we will not have to ask for it again?