Summary
Anthropic, a major artificial intelligence company, recently admitted that technical mistakes caused its coding tool to perform poorly for over a month. The tool, called Claude Code, had become very popular with software developers, but many users noticed it was getting worse. After weeks of denying there was a problem, Anthropic released a report explaining what went wrong. This situation has upset many customers and raised questions about whether the company can keep up with its fast growth.
Main Impact
The performance drop has damaged the trust between Anthropic and the developers who use its tools. For weeks, the company suggested that users were imagining the problems or that the changes were actually helpful. This led to accusations of "gaslighting," where a company makes customers doubt their own experiences. As a result, some users have canceled their paid subscriptions and moved to rival services like OpenAI. The event also highlights the struggle AI companies face in finding enough computer power to run their systems as more people sign up.
Key Details
What Happened
Anthropic explained that three specific engineering errors caused the decline in Claude Code. First, the company lowered the AI's "reasoning effort" to make it respond faster, but this made the AI less smart. Second, a technical bug caused the AI to forget what it was doing in the middle of a task. This made the tool act erratic and used up the customers' monthly limits much faster than usual. Finally, the company tried to limit the AI's responses to just 25 words at a time, which made the quality of the computer code it wrote much worse.
Important Numbers and Facts
Anthropic is currently valued at $380 billion, and its yearly revenue has jumped to $30 billion. Despite this financial success, technical experts found serious issues during the month of poor performance. One security firm reported that Claude’s code quality dropped by 47%. Another study found that the AI introduced security risks in more than half of the tasks it was given. To make up for these issues, Anthropic reset the usage limits for all its subscribers on April 23, 2026.
Background and Context
Anthropic was started by former employees of OpenAI. The company has always tried to show itself as a more honest and safety-focused alternative to its competitors. Developers liked Claude Code because it was often better at complex programming tasks than other AI models. However, the AI industry is currently facing a "compute crunch." This means there is not enough high-end computer hardware available to handle the massive amount of work these AI models require. When an AI tool becomes very popular very quickly, the company must find ways to save power, which can sometimes lead to lower quality.
Public or Industry Reaction
The reaction from the tech community has been mostly negative. Many developers expressed frustration on social media, saying the tool became "unusable" for difficult work. Some experts warned that the poor-quality code created by the AI could lead to dangerous security holes in software. Meanwhile, competitors have taken notice. OpenAI recently released a new model called GPT-5.5 and claimed that Anthropic made a mistake by not securing enough computer power early on. Some users feel that Anthropic only admitted to the mistakes because they were losing too many customers to rivals.
What This Means Going Forward
Anthropic is now trying to fix its reputation. The company has promised to be more transparent about any changes it makes to its software in the future. To solve the power shortage, Anthropic is expanding its partnerships with Google and Amazon to get more computer resources. However, the company is also testing new ways to manage costs. This includes trying out higher prices for heavy users and limiting access to certain features on cheaper plans. The main challenge for Anthropic will be proving that it can provide a stable, high-quality product while growing at such a high speed.
Final Take
This incident shows that even the most valuable AI companies can struggle when they grow too fast. While technical bugs can be fixed, losing the trust of a loyal community is much harder to repair. Anthropic must now balance its need for more computer power with its promise to provide the best tools for developers. If the company cannot maintain its quality, it risks losing its spot as a leader in the AI race.
Frequently Asked Questions
Why did Claude Code stop working well?
It was caused by three technical mistakes: reducing the AI's thinking power to save time, a bug that made the AI forget its history, and a rule that limited how much the AI could say at once.
Has Anthropic fixed the problems?
Yes, the company says all three technical issues were fixed by April 20. They also reset usage limits for subscribers to make up for the lost time and errors.
Is Anthropic running out of computer power?
The company admitted that demand has been very high and their systems have been stretched. They are currently working with Amazon and Google to add more computer capacity to handle the growth.
