This is the final part of the series inspired by Irina Nazarova's keynote "Startups on Rails" at RubyConf Thailand 2026. We've already covered psychology (Part 1), business cases (Part 2), and benchmarks (Part 3). Today — an honest conversation about weak spots and a plan of action.

Ruby is great for getting started, but if we want the stack to stay competitive in the AI era, we can't ignore the systemic gaps.

Notably, the founders Nazarova interviewed are not just building products — they're building open source tooling for the ecosystem. Carmine Paolino created RubyLLM (3,400+ stars). Kieran Klaassen built compound-engineering (6,200+ stars), a plugin for AI-assisted coding. Brad Gessler is building Terminalwire and Superform. They're doing this because something is missing. When startup founders spend time on infrastructure instead of their product, we should pay attention.

Here are the 3 key gaps that founders highlighted.

1. Infrastructure for Multi-Agent Systems

AI today is not just a request-response pattern. It's complex chains of agents that work asynchronously, wait for responses from third-party APIs, stream data, and switch contexts.

  • Problem: Brian Moseley, CTO of Sixfold ($21.5M raised), admits: "The agent's box is written in Python today. Not proud of that fact." He keeps business logic on Rails but is forced to build agent infrastructure on Python. The result is two monoliths in one company.
  • What's missing: Agent orchestration, built-in tracing, agent-to-agent protocols.
  • Solution: The Ruby ecosystem needs its own standardized, async tool — something like ActiveAgent (similar to ActionMailer or ActiveJob) — that becomes the standard for orchestrating AI agents.

2. Typing and Code Generation

The discussion around Sorbet and RBS continues. Evgeny Li, founder of Bemi AI, puts it bluntly: "Gradual typing in Ruby is a mess — fragmented between RBS, Steep, Sorbet."

The fact remains: large companies use typed Ruby extensively. Shopify has 99% of their 75,000 files typed.

For AI code generation, types provide a critical advantage. As Vladimir Dementyev said at his Rails X keynote: "Types are not just for safety. They're a source of information for AI tools." Types serve as hard constraints for neural networks, preventing them from generating non-existent methods. Settling the typing story is an important step for the ecosystem.

3. Delayed Official SDKs from Vendors

Remember the Anthropic (Claude) launch? The official Python and TypeScript SDKs shipped before launch.

What happened with Ruby?

  1. Alex Rudall built the anthropic gem himself.
  2. The gem became so popular that Anthropic came to him two years later asking to take over the gem name for official support.

From Anthropic's official SDK README (April 2025): "Thank you @alexrudall for donating the anthropic Ruby gem name and paving the way by building the first Anthropic Ruby SDK."

  [ New LLM / API release ]
             │
             ├──► Python / JS: Official SDK on day 1
             │
             └──► Ruby: Ecosystem waits for volunteers ──► Community gem ──► (1-2 years later) Vendor support

In Ruby, you build — and the giants come to you. But that means nothing happens "by default."

What founders don't complain about

This is worth noting. None of the 50+ founders Nazarova interviewed mentioned these as blockers:

  • "Ruby is too slow" — not a single mention of performance as a problem.
  • "ActiveRecord doesn't scale" — Intercom handles 2M req/sec; AngelList manages $100B.
  • "Can't hire Ruby developers" — hiring is harder (smaller pool), but the developers you find are more experienced. Plus there's an influx of developers switching from JavaScript.
  • "We need to rewrite" — zero migration plans. Zero.

The Red Pill for the Ruby Community: What to Do Right Now

Choosing Ruby in 2026 is not choosing the path of least resistance. It's a deliberate choice of a path where your contribution matters.

       How every engineer can help the ecosystem:

  1. Write open source (like Carmine with RubyLLM or Kieran with compound-engineering)
  2. Demand SDKs from vendors (tag ClickHouse, Anthropic, OpenAI)
  3. Don't hide your stack (write "Ruby/Rails" in job postings and blogs)
  4. Contribute and collaborate with open source authors
  5. Promote a constructive vision of Ruby's future
  6. Grow local meetups and conferences
  7. Celebrate Ruby success stories
  8. Bring people into the community

"Rails is not a single perfect idea. It's a messy quilt that we are constructing together." — Vladimir Dementyev, Rails X keynote, SF Ruby Conference

The question isn't "Will Ruby survive?" It's: "What will you build?"

Reading checklist for the full series:

Have thoughts or feedback on this dispatch? GET IN TOUCH →