June 2025 Summaries

Trophy, a developer-focused gamification stack, transitioned from using Fern to Mintlify for their API documentation, driven by the need to avoid vendor lock-in associated with Fern's proprietary syntax and to improve visual design. Mintlify was chosen for its intuitive MDX workflow, OpenAPI-first approach, and rich component library, which seamlessly integrated with Trophy's existing processes and significantly enhanced the developer experience. The migration process was smooth due to Mintlify's support for Lucide icons and a 'link rot' CLI tool that checks for broken links, along with excellent customer support that addressed any issues promptly. This transition allowed Trophy to create documentation that aligns closely with their product, ensuring developers can quickly understand and implement their API features. Looking forward, Trophy is eager to collaborate with Mintlify on future innovations, including AI-driven features and multi-language support, reaffirming their commitment to exceptional developer experiences.

A recent analysis by Profound indicates that the llms-full.txt file receives significantly more AI traffic compared to llms.txt, with ChatGPT being the primary visitor. This preference is attributed to LLMs' inclination to embed comprehensive content upfront, as opposed to using retrieval-augmented generation, which can be hampered by retrieval latency and inconsistent formatting. The study, which examined traffic from 25 companies over a week, found that llms-full.txt provides a richer, more structured dataset, allowing for more efficient indexing and retrieval by AI models like GPT-4-turbo. This format, containing substantially more content than llms.txt, aligns well with large context windows and offers a more streamlined approach to indexing, reducing fragmentation and enhancing retrieval accuracy. While embedding llms-full.txt might initially be more resource-intensive, it ultimately facilitates faster and more consistent AI responses from cached embeddings, highlighting its growing importance in making documentation AI-ready.

Replit has significantly revamped its documentation to align with its shift towards AI, aiming to cater to "vibe coders" and AI agents while establishing scalable processes such as Vale linting and AI-assisted workflows. This transformation reduced manual changelog efforts by 90% and created a dynamic documentation system that evolves with their fast-growing product. The overhaul was necessary as Replit's existing documentation was outdated for their new AI-focused direction. They reorganized the content to align with user goals and adopted the Diátaxis framework for better navigability and discoverability. To maintain quality over time, Replit implemented systems like pre-commit hooks and Markdown-based style guides, and embraced AI tools to expedite writing processes, allowing AI to draft content while humans refine it. This approach not only streamlined documentation updates but also made it easier for team members to contribute, turning their documentation into a scalable support system that enhances user success.

The guide offers a detailed approach to creating effective API documentation that is designed to be intuitive for both human developers and large language models (LLMs), using examples from notable companies like Stripe, Twilio, and Pinecone. It highlights the importance of clear, structured documentation that facilitates a smooth developer experience, emphasizing the role of documentation in driving API adoption and reducing the support burden. Common pitfalls such as context mismatch, outdated content, and documentation locked behind authentication walls are discussed, alongside strategies for overcoming them. The guide underscores the need for comprehensive API references, practical guides, visible changelogs, and effective error handling to enhance usability and trust. By treating documentation as a product, with assigned ownership and regular audits, developers can ensure their APIs are easily understandable and usable, ultimately leading to faster adoption and fewer support issues.

Claude, a Large Language Model, utilizes two distinct context management methods: memory and the Model Context Protocol (MCP), each serving different purposes to enhance user interactions. Memory allows Claude to retain user preferences and context across sessions, thus providing continuity by remembering details such as preferred programming languages and response styles. This persistence makes it ideal for long-term projects where continuity is essential. In contrast, MCP offers real-time access to external data sources like documentation and APIs, ensuring responses are current by retrieving specific information during a session. This makes MCP suitable for scenarios requiring up-to-date data, like customer support involving frequently updated API documentation. For optimal performance, both memory and MCP rely on well-structured documentation, which should include clear hierarchies, metadata, and consistent formatting to facilitate accurate and efficient data retrieval. By combining memory and MCP, Claude delivers personalized and precise responses, contingent on the quality and structure of the provided documentation.

Mintlify's new AI Assistant enhances user interaction with documentation by utilizing agentic retrieval technology to deliver more accurate, conversational responses rooted in a company's source of truth. This AI-native experience improves user engagement by understanding intent and providing contextual answers with citations, while maintaining control over product information presentation in AI interfaces. The assistant leverages advanced models like Claude 4 for high-quality answers with minimal hallucination, even for complex queries. It offers a seamless, embedded experience that allows users to interact directly within documentation, including querying from code blocks for instant explanations. By integrating AI functionality into documentation, companies can provide trusted, fast responses and gain insights into user queries, ultimately allowing them to maintain narrative control and improve content accuracy. Mintlify is also focused on automating the upkeep of documentation by identifying knowledge gaps and facilitating updates through AI, with examples of the assistant's application seen in companies like Perplexity and Bolt.new.

This guide reviews the top API documentation tools of 2025, categorizing them into end-to-end platforms, API reference generators, tools with embedded documentation, and static site generators. Each tool is evaluated based on its ability to auto-generate documentation, facilitate live collaboration, integrate with Git workflows, offer robust search functionality, and enhance the developer experience. The evaluation aims to assist teams in selecting the most suitable documentation tools that align with their workflows, whether prioritizing automation, customization, or collaboration. Key tools discussed include Mintlify, which balances automation and customization while providing interactive features and AI-powered search; ReadMe, known for its interactive onboarding and collaboration features; GitBook, which emphasizes ease of use for non-technical users; and Docusaurus, which offers high customization for developers familiar with React. Additionally, the guide covers API reference generators such as Swagger UI and Redoc, highlighting their strengths in creating clear, interactive API references. Lastly, it explores tools with embedded documentation capabilities like Postman and Insomnia, as well as static site generators like MkDocs, emphasizing their suitability for different team requirements and technical expertise levels.

Emma Adler's article outlines a three-part formula for creating effective developer documentation: clarity, context, and flow. It emphasizes the importance of treating documentation like a well-designed storefront, where everything is intuitively laid out to guide developers seamlessly from browsing to taking meaningful actions, such as making their first API call or integrating SDKs. The article highlights that many developers abandon documentation because of unclear pathways and lack of context, leading to missed conversion opportunities. It advises against treating documentation as a warehouse of information and instead suggests using strategies like progressive disclosure, visual hierarchy, and motivational framing to maintain developer engagement. Adler also discusses the significance of using the right documentation tools that support a product-focused mindset, such as those enabling version control, CI/CD integration, and accommodating various content types. By redesigning documentation with these principles, developers can have a more efficient and satisfying user experience, ultimately leading to greater product integration and advocacy.

Building a startup is portrayed as a long-term commitment rather than a race, emphasizing the importance of perseverance through challenging times. Co-founder Han Wang shares personal experiences of facing struggles, such as enduring multiple pivots and feeling pressure from peers' rapid successes, to illustrate the unseen hard work behind the scenes. A pivotal moment came from a successful founder's story, highlighting patience and persistence over five years before achieving their first customer. This narrative underlines the significance of staying grounded and committed, reminding founders that the journey is about enduring the unseen years and maintaining belief in eventual success. Wang encourages sharing both the highs and lows of startup development to inspire others to persist through similar challenges.