David Heinemeier Hansson announced at Rails World 2026 that he has retired from professional programming and now relies on LLM code generation, claiming humans reading code should be rare. He revealed that 37signals is moving Hey away from Rails to native applications built with LLMs and Rust on the backend, fundamentally shifting his vision of Rails from a framework for ambitious small teams to primarily a platform for web apps of necessity.
Rails' convention-over-configuration approach makes it exceptionally suited for AI coding agents, providing clear structural patterns that reduce token usage and improve code generation quality. Benchmarks comparing AI models on Rails feature tickets and API tasks show that models perform better with Rails' standardized conventions, enabling solo developers to ship faster with AI assistance.
Topcoat, a batteries-included Rust web framework, has released version 0.9 with client-side reactivity features. The framework aims to combine server-side and browser-side rendering to provide productivity comparable to Rails while leveraging Rust's performance and safety. Created by developers with deep experience in both Rails and Rust's networking ecosystem, Topcoat targets AI-driven development workflows.
Programming tutorials are losing market relevance as AI coding agents eliminate the traditional translation step between sample applications and real-world codebases. Education business revenue is declining, with established creators like GoRails reporting AI's impact on sustainability, shifting how developers learn to build features in their actual applications rather than abstract examples.
A Rails 8.1 photo gallery app (friendlyfolio) serving 1,916 photos from three Friendly.rb conferences replaced a hosted service. The team optimized image delivery using Active Storage variants in AVIF, WebP, and JPEG formats across multiple sizes, learning key lessons about image optimization at scale that go beyond basic tutorials.
A Ruby gem that integrates AI-powered judgment into code by adding System One (S1) model capabilities as native features. The gem enables typed probabilistic queries about data without generation, allowing code to ask questions and make decisions based on calibrated probability responses.
A programming language design discussion comparing Rust's minimal parameter features with Ruby's flexible argument handling, examining named parameters, defaults, and keyword arguments through examples like Rails's redirect_to function and FastAPI decorators.
RubyLLM 2.0 expands the framework with new provider coverage (now seventeen total), AI operations including tool approvals and citations, conversation controls for agentic workflows, and Rails integration. The release adds support for video, speech, OCR, reranking, batches, and usage tracking while maintaining consistent APIs across providers.
RubyLLM 2.0 expands support across multiple AI providers and introduces new features including video, speech, OCR, batches, and Rails integration. The framework adds human approval controls for tools, citations for document and web references, and separates providers from protocols for consistent APIs across different services.
RubyLLM 2.0 expands the framework with support for new providers (Cohere, Deepgram, ElevenLabs, Ollama Cloud), AI operations (video, speech, OCR, reranking, batches), and conversation controls including tool approvals, citations, and agentic workflows. The update adds consistent Ruby APIs across seventeen providers while maintaining streaming and usage tracking throughout.