What About Rails?
Rails developers are grappling with the framework’s future after its creator used a keynote to champion AI‑driven development and a move from Rails to native Rust applications for his own products. Many argue that large language models erase Rails’ traditional advantages in developer productivity and make statically typed, high‑performance stacks more attractive, while others counter that Rails remains mature, viable, and productive for CRUD and business apps. The debate widens into questions about overreliance on a single project leader, whether web frameworks and UIs will be displaced by agents and CLIs, and what AI‑generated code means for maintainability, performance, and developer careers.
Reaction to the keynote and “Rails is done”
- Many felt the keynote at a Rails conference barely addressed Rails’ concrete future, instead centering on AI agents, Rust, and native apps.
- Attendees were frustrated that the flagship product is moving off Rails and that the message to existing users was basically “you’re the best, but I’m leaving.”
- Some see the talk as honest and timely: Rails solved “don’t write boilerplate” in 2004; AI now solves that better, so it’s rational to move on.
- Others think declaring Rails “over” is premature or self‑defeating, especially while people run real businesses on it with good performance.
Rails’ future, forks, and BDFL dynamics
- Several argue Rails is tightly bound to its creator’s personality; when he pivots, the community is left in limbo.
- Others counter that Rails has many maintainers and ongoing commits, so it can continue as a mature, stable framework.
- A fork (e.g., Mosscap) is discussed; people question whether it will meaningfully ship or is mostly political branding.
- Some explicitly call for a “hard break” from the original leadership; others say critics are entitled and can always contribute or fork.
AI, languages, and frameworks
- A strong camp claims dynamic languages (Ruby, PHP, Python) are becoming obsolete: with LLMs writing code, you can use statically typed, fast, memory‑safe stacks (Rust, Go, etc.) with little extra effort.
- Counterargument: frameworks like Rails/Laravel still compress domain concepts into well‑known patterns, making them easier for both humans and LLMs to work with.
- Disagreement over performance: some insist Ruby/PHP are IO‑bound and inherently slow; others point to benchmarks and note many apps are DB/network‑bound where Rails is “fast enough.”
Web vs native; UI vs chatbots
- The keynote’s “native apps + CLI/agents” direction sparked debate:
- Some see AI‑built native apps as the logical successor to web apps and Electron.
- Many push back: good UIs still matter, chatbots are often worse UX than GUI or CLI, and “nobody wants to use a chatbot for everything.”
- There’s interest in hybrid models: build strong CLIs/APIs for agents, but keep human‑friendly UIs on top.
LLM‑written code and engineering practice
- Some report successful revenue‑generating apps where 80–90% of the code is LLM‑generated, with humans focusing on tests, architecture, and product decisions.
- Others warn that agents happily produce inefficient, unscalable, or messy designs unless tightly constrained and reviewed.
- A recurring concern: if devs “stop reading code,” comprehension and maintainability may degrade; others reply that tests and agents’ code explanations offset this.