Quota Crisis, Day 2

Twenty-four hours after we published The Quota Crisis, nothing has changed — except that more users are reporting the same problem. Anthropic has not issued a public statement. The company's status page still shows Claude Code as fully operational. GitHub issues continue to be closed as "invalid" or "duplicate." And a temporary promotion that doubled off-peak limits is set to expire on March 28 — three days from now.

New Reports, Same Symptoms

GitHub issue #38701, filed today, reports "unexpected rapid consumption of plan usage quota" on a paid plan exceeding €100/month. The user says they previously never hit their limit. Now they run out within an hour. A commenter confirmed identical symptoms: "Current session 5h window usage jumped to 25% without even starting a single conversation. This is happening consistently for a past day or so."¹

Issue #38369, filed March 24, reports "subscription usage jumping to 80% within minutes of reset."² Issue #38029 documents abnormal quota consumption specifically on session resume — the act of reconnecting to a previous conversation drains quota disproportionately.²

The total count of quota-related issues filed since March 23 now stands at at least 10: #38701, #38369, #38357, #38350, #38344, #38335, #38330, #38239, #38064, and #38029.² Anthropic's response pattern remains consistent: issues are labeled "invalid" (classified as API-level, not Claude Code), closed as duplicates, or left open without comment.

The Promotion That Expires Friday

On March 13, Anthropic quietly launched a temporary promotion: 2x usage limits during off-peak hours, running through March 28. Off-peak doubled limits do not count toward weekly caps.³

This matters because the promotion may be masking the true severity of the quota reduction for some users. Developers working evenings and weekends — when off-peak hours apply — are getting temporarily doubled limits that soften the blow. But the underlying quotas remain reduced from pre-March levels.

On March 28, the promotion ends. If quotas revert to the reduced baseline without a new plan, pricing change, or public acknowledgment, every Max plan user will feel the full impact simultaneously. Friday could be the real crisis.

The Burn Rate Variance

The most concerning technical detail is the inconsistency. The forensic analysis in our original report showed one user's quota draining at a predictable (if accelerated) rate. But subsequent reports show wild variance.

One user documented burn rates varying from 5.6% per hour to 59.9% per hour on the same plan within a 48-hour window — an order of magnitude difference with no change in usage patterns.⁴ This suggests the quota calculation isn't just stricter than before — it may be inconsistent, which is harder for users to plan around and harder for Anthropic to explain as an intentional change.

The Bigger Picture

Anthropic is under unusual financial pressure. On March 5, the Pentagon formally designated Anthropic a "supply chain risk" — the first time a U.S. AI company has received this designation — after Anthropic refused to remove restrictions on mass surveillance and autonomous weapons use cases.⁵ OpenAI took the defense contract. Anthropic sued. On March 24, a federal judge said the designation "looks like an attempt to cripple" the company.⁶

We are not claiming the Pentagon situation caused the quota changes. The timeline is suggestive — the designation took effect in early March, the quota anomalies began around March 22-23 — but correlation is not causation, and Anthropic may have entirely separate reasons for adjusting how tokens are weighted against plan budgets. What we can say is that a company under financial pressure from a lost defense contract has economic incentives to tighten the unit economics of its consumer products.

One Hacker News commenter put the pricing dynamic bluntly: "They were selling you $1,000 worth of tokens for $200, and now they're cracking down."⁷ If that's accurate — if Max plans were consistently unprofitable — the quota reduction may be a correction rather than a bug. But corrections deserve announcements, not silence.

What We're Watching

March 28: The 2x off-peak promotion expires. If limits revert to the reduced baseline with no announcement, that confirms the reduction is intentional and permanent. If Anthropic extends the promotion or adjusts pricing, that's a different signal.

GitHub issues: The volume and specificity of new reports. If the "invalid" labels continue with no engineering response, the community will draw its own conclusions.

An official statement: Still waiting. Every day without one is a day the community fills the void with speculation. Bug or feature, silence is the worst response.