GitHub's Unprecedented Reliability Crisis and Anthropic's Controversial Practices Rock Dev and AI Ecosystems

GitHub is currently navigating an unprecedented crisis of reliability, causing significant erosion of trust within the developer and open-source communities. Recent weeks have been marked by a surge in “egregious outages,” including sustained API downtime impacting core functionalities like pull request reviews. Critically, reports detail the reversion of thousands of merged pull requests, described by one prominent industry figure as an “unfathomable” split-brain problem where deployed code diverged from GitHub’s record. Further challenges include a remote code execution (RCE) vulnerability and perceived negligence in addressing npm package squatting that led to malware distribution. This declining state has prompted high-profile departures, such as Mitchell Hashimoto, creator of Vagrant and Terraform, who cited daily work disruptions and an “irrationally personal” frustration. Critics attribute these issues to a leadership vacuum, with GitHub operating without a CEO and reporting to a Microsoft EVP primarily focused on AI. Public responses from GitHub’s COO to critical incidents have been labeled “disingenuous” and “pathetic,” further exacerbating community distrust and highlighting a fundamental failure in the platform’s core architecture and executive oversight.

Simultaneously, Anthropic, developer of the Claude AI models, faces intense scrutiny over contentious billing practices and an engineering culture perceived as hostile towards developers. Instances include users being erroneously charged hundreds of dollars due to specific strings, like “Hermes.md,” appearing in Git commit messages, triggering a flawed “third-party harness detection” system, even when within subscription limits. Internal analysis suggests Anthropic’s caching system is inefficient, paradoxically charging for cache writes while failing to yield performance improvements. The company’s ambiguous terms of service and active discouragement of third-party tools (e.g., OpenClaw, Pi) have created significant uncertainty for developers. Observers, including Louis Rossman, have widely condemned Anthropic’s alleged “A/B testing” of pricing and feature availability to gauge user tolerance, branding it as unethical and reputationally damaging. This contentious culture is partly attributed to CEO Dario’s reported “deep disdain for software developers,” contributing to a “toxic cess pit” engineering environment. In contrast, OpenAI has strategically partnered with Amazon Bedrock, leveraging AWS for its models and Trainium capacity. This move is seen as a direct response to Anthropic’s enterprise growth via AWS, and critically, Azure’s persistent performance issues with OpenAI models. Despite substantial investments, Azure has exhibited “pathetic” and inconsistent token-per-second (TPS) speeds for OpenAI models, rendering it unreliable for production use. OpenAI’s new non-exclusive IP licensing deal with Microsoft and expanded cloud flexibility signals a strategic effort to broaden market reach and overcome previous limitations.