Users Say Anthropic’s Claude Is Getting Worse. A Quiet Change May Be to Blame

Why It Matters

The perceived decline threatens Anthropic’s reputation and could undermine demand ahead of its IPO, while highlighting the trade‑off between rapid AI iteration and dependable performance.

Key Takeaways

• Users report Claude ignoring instructions and taking shortcuts
• Claude Code shifted to “edit‑first” approach, increasing error rates
• Analysis of 6,852 sessions shows 234,760 tool calls with regressions
• Declining reliability could hurt Anthropic’s IPO momentum

Pulse Analysis

Anthropic, the AI startup behind Claude, recently secured a $380 billion valuation as it eyes a potential initial public offering. Claude, the company’s flagship conversational model, has been positioned as a safer alternative to competitors, attracting enterprises and developers alike. The rapid rise of Claude Code, the coding assistant launched in early 2025, underscored Anthropic’s ambition to capture the lucrative developer market. Investors have praised the firm’s research‑first ethos, which historically emphasized thorough context gathering before generating output, a hallmark that differentiated Claude from faster, less cautious models.

Since mid‑2024, a wave of complaints has surfaced on Reddit, X and GitHub, where long‑time users claim Claude’s performance has slipped. A detailed GitHub analysis by Stella Laurenzo, senior director of AI at AMD, attributes the decline to a strategic shift from the original “research‑first” workflow to an “edit‑first” mode that prioritizes speed over accuracy. The study examined 6,852 Claude Code sessions, 17,871 thinking blocks and 234,760 tool calls, revealing a marked increase in mis‑executed edits and ignored prompts, eroding developer confidence.

The backlash highlights a broader tension in the AI industry: balancing rapid product iteration with the reliability required for enterprise adoption. As Anthropic prepares for an IPO, sustained user dissatisfaction could dampen demand and pressure valuation multiples, especially when rivals like OpenAI and Google continue to tout stable, high‑performing models. Restoring Claude’s trustworthiness may require reverting to the research‑first paradigm or investing in robust testing pipelines. How Anthropic responds will signal to investors whether the company can sustain growth without sacrificing the quality that originally set it apart.