August 2025 Summaries

The concept of llms.txt, likened to a robots.txt for AI, has been critiqued as ineffective and overhyped, with tests showing it requires manual input into language models to function, rather than being spontaneously recognized or respected by them. Redocly experimented with llms.txt, finding it offered little value, as logs indicated negligible access or systematic use by AI models. The exploration led to a more practical development: a one-click feature to copy pages in Markdown for seamless integration with AI systems like ChatGPT, which proved more beneficial. In contrast, another project involving new MCP features demonstrated notable utility within AI workflows, highlighting the disparity between the limited impact of llms.txt and the promising potential of effective content integration tools. The narrative suggests that meaningful content governance in AI will depend on robust licensing, attribution, and legal standards, rather than underutilized text files.

Supporting multiple API versions simultaneously can increase costs and operational overhead, as each version introduces additional complexities in terms of support, infrastructure, and maintenance. API versioning strategies typically fall into two main approaches: evolution and explicit versioning. Evolution minimizes version proliferation by maintaining a single API version and making non-breaking updates, while explicit versioning creates distinct versions, offering clearer boundaries for modifications. Many successful APIs, such as Stripe, use a hybrid approach that combines these strategies to balance stability and flexibility. Effective versioning requires planning for version retirement, with clear deprecation timelines to minimize long-term maintenance burdens. OpenAPI specifications provide a structured method for managing versioned APIs, supporting both evolution and explicit strategies, and enabling automation of maintenance tasks. Clear communication with API consumers is crucial for successful versioning, involving systematic announcements and guidance to ensure smooth transitions. Additionally, documentation platforms must support multiple API versions to provide users with the necessary reference materials and facilitate version switching without adding administrative overhead.

During the summer months of 2025, Redocly introduced a range of new features, updates, and enhancements across its product suite, including Realm, Reef, Revel, and Reunite, with notable improvements in internationalization, accessibility, AI integration, and analytics. Among the key updates, version 2.0 of the Redocly CLI was released, streamlining the command-line interface with modernized features, enhanced security, and validation capabilities. The company also improved its AsyncAPI documentation, enriched developer experiences with new hooks and compatibility features, and provided new formats for documentation, such as code groups and request/response samples. Quality of life enhancements were made, such as path parameter highlighting in request URLs and branch deployments in Reunite. Redocly is actively seeking software engineers to join their team, as they continue to invest in future developments like MCP servers, runtime logs, and visual workflow builders to further enhance API documentation and developer tools.

In an evolving digital landscape where developers increasingly rely on AI agents to navigate documentation, optimizing documents for large language models (LLMs) is crucial for accurately conveying API information. Many developers face challenges due to unoptimized documentation, which leads to incorrect AI-generated responses, support ticket inflation, and onboarding friction. The text outlines strategies to enhance LLM compatibility by maintaining a predictable heading structure, using consistent terminology, properly formatting code examples, providing text alternatives for visual content, and being explicit in descriptions. Addressing semantic drift and periodically auditing documentation for consistency are recommended to prevent misinformation. Testing documentation with AI tools and focusing on high-impact content can ensure improved accuracy and efficiency. Platforms like Redocly, which generate documentation from OpenAPI specs, are highlighted as effective solutions for creating AI-friendly documentation without sacrificing human readability.