Anthropic Faces Backlash Over Misleading Usage Limit Changes

Anthropic announced a permanent 25% increase in standard weekly limits for Pro, Team, and seat-based Enterprise plans, effective September 14th. This update comes as a replacement for a prior temporary 50% increase that was previously in effect for users.

Theo criticized Anthropic for poor transparency and communication surrounding these updates. The original announcement post on social media was deleted and replaced with a new version, leading to confusion among users who found the initial screenshot to be cut off from key information.

Theo shared his extensive personal experience with Anthropic's rate limits, noting he maintains five separate accounts to manage his usage. He reported frequently hitting usage caps in the past, asserting deep familiarity with Anthropic's scaling and compute issues.

Theo expressed frustration, stating a new model was expected two days prior to the post. He argued that the ongoing compute constraints are significantly hindering his ability to code efficiently. Additionally, he criticized the timing of the announcement, noting it occurred on a Saturday morning.

PostHog is transforming its all-in-one suite of product tools into a self-driving product iteration loop. This tool uses analytics to automatically identify and fix user-facing issues, with 97% of PostHog users currently paying nothing due to its generous free tier.

Users are actively mocking Anthropic's announcement on social media platforms, with Theo highlighting phrases such as 'Stop hitting yourself' as a common sentiment. This reflects widespread dissatisfaction with the company's policy changes.

In May, Anthropic announced a partnership with SpaceX intended to substantially increase its compute capacity, which was stated to enable increased usage limits for Claude Code and the Claude API. Initially, Anthropic doubled the 5-hour rate limits for Claude Code, but Theo noted that the weekly limit, not the 5-hour limit, was the actual persistent issue for users.

In May, weekly limits were temporarily increased by 50% until July 13th. The removal and subsequent re-addition of the 'Fable' model caused significant user frustration due to its perceived superior performance. Anthropic then restricted Fable usage to 50% of the total weekly limit, effectively forcing users to switch to other models like Opus or Sonnet to conserve their Fable allocation.

Theo argued that the current Opus and Sonnet models underperform compared to Fable. He explained that the 50% Fable limit policy effectively cuts usable capacity in half for heavy Fable users, as many exhaust their Fable allocation quickly and are left with less desirable alternatives. Theo also expressed regret for previously praising Opus as a high-quality model, noting it has become a meme that he has virtually unlimited Opus usage because his multiple accounts hit their Fable limits.

Theo criticized Anthropic for framing a 17% reduction in effective limits as an increase, calling it "the worst way of saying your limits are going down by pretending that they're actually going up." Many heavy users had joined or upgraded during the promotional 50% increase period, only to find themselves with a poor experience due to the impending 17% cut they were not clearly informed about. Theo argued that this strategy effectively misinforms customers.

Theo proposed a more transparent approach to handling limit changes, suggesting companies should lead with bad news, be realistic, and acknowledge user impact directly. He provided an example of empathetic communication: "We know that this will impact the work of many users. We've run the numbers and found that we were able to confidently make a permanent increase of 25% from the original base. This nets out to a 17% decrease from the limits as they are today."

Theo noted that Anthropic communicates clearly only when it serves their competitive interests, criticizing their response to OpenAI and Cursor's relationship changes. He alleged that Anthropic views customers merely as a 'means to an end,' primarily focusing on AGI development rather than user support.

Theo suggested that Claude might be used to draft communications with specific instructions to always sound positive, even when delivering bad news. This approach, he argued, leads to an inability to properly acknowledge legitimate user pain points and fosters a corporate tendency to frame negative service changes as positive developments.

Theo critiqued Anthropic's language as insincere and lacking real empathy, arguing that the company avoids admitting fault when issues arise. He accused the company of blaming 'compute' or external factors rather than taking responsibility for their decisions.

Theo argued that subscription-based access now acts as a 'loss leader' compared to direct API sales, implying that Anthropic prioritizes enterprise customers who pay per usage with higher margins. He stated that during peak business hours, compute scarcity leads to failed requests for subscribers in order to protect enterprise profit.

Theo criticized the combination of reduced weekly limits and stricter constraints as a clear downgrade that Anthropic is pretending is not. He expressed hope for the upcoming 'Astro' model to offer a better alternative and described arbitrary restrictions as a signature 'Claude special' that frustrates users.

Theo urged Anthropic to stop lying to their community, stating that their generally anti-user stance could ultimately destroy them. He compared Anthropic's strictness with OpenAI's more flexible approach, warning that this behavior could cost them their market position as competitors catch up.

Theo pleaded with Anthropic to stop self-sabotaging their reputation, asking them to "stop hitting yourselves." He expressed a desire to continue defending the product quality without having to defend the company's business practices.