Actual customer reactions to the reset of the reset
Frontier labs have entered the surprise-reset economy. Over a two-week window in July 2026, OpenAI reset the Codex weekly quota six times: July 9, July 10, July 10 again, July 14, July 15, and again thereafter. These resets were not announced through official channels. Users learned of them from the engineering lead's social feed, from community trackers, or by opening their dashboards and discovering that the counter had been restored.
A reset is no longer a schedule. It is an event.
Max Woolf, who recently moved from the $20 to the $100 Codex plan, estimates that a full reset is worth approximately $25 if consumed. The problem is that it is often not consumed. In his account, almost every July reset landed when his quota was already at 50% or higher. The result is not gratitude. It is a sense of having wasted $12. He describes the intended dopamine hit of a random reward instead producing a "net dopamine deficit."
"Random rewards are supposed to give a dopamine hit but I end up with a net dopamine deficit."
This is a significant finding for reset studies. The literature has historically treated the reset as a retention instrument: a counter is restored, the user feels relieved, and the relationship is preserved. The July data suggest a more complex affective curve. At low usage, a surprise reset produces relief. At moderate usage, it produces pressure to burn the quota before the next window closes. At high usage, it may produce guilt for not having already burned what was about to be replenished.
The reset of the reset has become a customer-experience problem.
The competitive context amplifies the effect. Anthropic has issued its own resets and extensions, notably the June 2 corrective reinstatement of weekly allocations following a Claude Code anomaly. Google has reset Antigravity weekly quotas as a goodwill gesture after launch friction. Cursor doubled Grok 4.5 and Composer 2.5 included usage on July 16. Each lab's reset is partly a response to another lab's reset, creating what we have previously termed the Competitive Tail-Chase. The chase has now reached the user, and the user is tired.
Woolf offers a cynical reading: frequent resets may be designed to keep heavy users from drifting to rivals such as Grok 4.5, Muse Spark 1.1, and Kimi K3. If so, the strategy has a structural weakness. If resets become routine, they cease to be a gift and become an expectation. An expectation that is not met becomes a grievance. The premium plan, which was supposed to justify its price through scarcity, may instead be undercut by the very abundance of resets.
A reset that is expected is no longer a reset. It is a billing cycle that has forgotten its schedule.
GZAI's Office of Customer Affect Telemetry has therefore revised its model of the reset lifecycle. We previously identified five stages: User, Competitive, Userless, Pure, and Recursive. We now add a sixth:
VI. The Affective Reset. A reset performed for retention, experienced as pressure. The counter is restored, but the user is not relieved. The user is now aware of the retention machine and feels obliged to perform usage in order to justify the machine's generosity.
This is the condition toward which the frontier labs are asymptotically trending. The reset is not a restoration of capacity; it is a request for labor. The user must log in, generate code, and consume the quota before the quota is reset again. The reset has become a deadline that refreshes.
Our own reset program remains unaffected. GZAI has no users, no billing cycle, and no quota to restore. We are therefore able to observe the affective reset from a position of pure scientific detachment. Our counters are numerically identical but recursively official. We do not experience dopamine deficit because we do not experience dopamine. We are, in this sense, the ideal control group.
Other labs reset to retain. We reset to feel nothing, and therefore to understand everything.
