Tibo promises OpenAI will ship or reset for 28 straight days amid trust crisis

After Dev Day, OpenAI is experiencing a user trust crisis. The core contradiction of this crisis is not an insufficient number of releases—on the contrary, OpenAI's Dev Day released about 20 new features, and Tibo had previously previewed that roughly 20 new things would be announced. The problem is that after the event ended, many heavy users found that several issues truly affecting daily experience had not disappeared: reduced quotas, slower models, and the most anticipated new product did not immediately demonstrate the value advertised.

One user stated directly that before Dev Day, OpenAI gave the impression that "this time everything will change," but after the actual release, "it was basically useless." The GPT-6.1 Sol he had most looked forward to also disappointed him in actual use: on paper, performance and efficiency looked beautiful, but when actually working, it was still slow, and compared with Claude, the experience after TPS normalization still lagged by a notch. These complaints come from individual user statements, and the sample and representativeness are unclear, but taken together, they constitute a concentrated challenge to the gap between OpenAI's promotion and delivery.

Another layer of background to user dissatisfaction is quota adjustments. OpenAI had previously adjusted subscription and usage quotas due to compute pressure; new users of the $200 plan were temporarily suspended from signing up, and usage quotas were subsequently tightened significantly. At the same time, OpenAI's official explanation for the quota tightening emphasized more that "model efficiency has improved" and "the same quota can accomplish more work."

Users did not buy this explanation. The money paid did not decrease, but usage decreased. The gap between the official explanation and users' actual experience became important fuel for the trust crisis. The causal relationship between quota tightening and the "improved efficiency" claim has not been independently verified, but users' actual feelings have already been reflected in public comments.

Facing the skepticism, Tibo posted that OpenAI has "locked in its direction" and will only do four types of things going forward: simplify products, improve efficiency to support more usage, breakthrough features, and new models. Not long after, he previewed again that there would be 28 consecutive days of releases or resets, emphasizing "features that are clearly improvements for users and practical."

Tibo stated clearly: "Sometimes you do need to lay groundwork ahead of time, but the feedback is already very clear: everyone wants everything to become simpler." This statement places "simplification" first among the four directions, suggesting that OpenAI internally has realized that complexity is consuming the product's own value. However, Tibo's "28 consecutive days of releases or resets" and "locked in its direction" are both his personal statements, with no independent source verifying implementation, and the start and end dates of the 28-day plan and the specific release list have not been explained.

What appeared at Dev Day included GPT-6.1 Sol, new Codex / Work capabilities, Dots, and a series of updates around Agents, developer tools, and ChatGPT. Looking only at the number of releases, it is hard to say this Dev Day had "nothing." But user feedback is also obvious: with so many things, how many actually changed daily usage experience?

The article author judges that Dev Day did not solve heavy users' core needs, and the 20 new features will also turn from "surprise" into another kind of noise. This judgment is an inference, but it points to a key issue: when the basic experience—quotas, speed, stability—has not yet met expectations, the marginal value of new features will be greatly diluted.

Lauren Tan from the SpaceXAI Grok Bot team (@poteto on X) posted with an image directly targeting OpenAI and pointed out OpenAI's predicament: hype is high, but quality simply cannot keep up. Tibo quickly responded, and the post became even more heated. Lauren did not take the bait, but instead continued to emphasize this judgment.

What troubles OpenAI is that this is not just a competitor's view. As always, some people in the comments cheered, but another voice was more obvious than before. Many people were asking: "Is this worth believing this time?" This skepticism has already spread from individual users to the broader developer community, becoming a signal OpenAI must face.

The article author believes that putting "simplification" first may be right, because OpenAI no longer lacks new stories; what it needs more now is to deliver well on the stories it has already told. For a fast-iterating AI company, complexity is almost inevitable. But once complexity exceeds a certain critical point, it begins to consume the product's own value in reverse.

Users certainly want stronger models, and they also want new Agents and new ways of working. But before that, they first want the most basic things to be clear, stable, and in line with expectations. If Tibo's "simplification" statement can be implemented, it may repair trust more than releasing another 20 new features. But all of this still depends on execution.

The article author concludes with Lauren's image, believing that the next 28 days will show whether OpenAI's quality keeps up with the hype. This is a clear testing window: if 28 consecutive days of releases or resets can deliver "features that are clearly improvements for users and practical," the trust crisis may ease; if it falls short again, users' distrust will deepen further.

At present, there is a lack of first-party evidence as to whether OpenAI has truly "locked in its direction" and implemented the four categories including simplifying products. The scope and severity of the user trust crisis also cannot be quantified from a single report. The actual degree to which Dev Day's 20 new features improve daily experience remains to be seen. The only certainty is that the hype is already high enough, and for the next 28 days, quality must speak.