Rails World 2026 Opening Keynote

Summary of the Rails World 2026 Opening Keynote (Tech + Product + Analysis)

Event / Livestream Context

• The opening keynote event in Austin includes a pre-show live stream, described as a first for Rails World.
• Shopify is the official event partner, supporting:
  • the livestream
  • additional programming such as a fireside chat sponsorship
  • a sponsor lounge
  • co-hosting the closing party

---

“Agentic Development” Focus (Recurring Theme)

Speakers repeatedly emphasize a shift away from traditional workflows (writing code directly) toward:

• AI agents that can execute tasks and produce software outcomes
• New integration patterns for AI within Rails applications
• Concerns to address, including:
  • cost
  • security
  • evaluation / benchmarking of agents

A notable framing was “prompt to IPO”:

• More teams are starting with prompts (agent workflows) rather than beginning with terminal-based development.

---

Segment 1: Matt / Matz Matsumoto — Ruby + AI Agent Coding

Key Points

• Matt says he uses AI extensively, including success stories around AI agent coding (e.g., “expert web coding”).
• He plans to discuss why some agent coding works despite “vibe coding” failures.
• He notes model quality improvements when prompting in Japanese, suggesting better multilingual capability improves outputs.
• He argues Ruby may be well-suited for AI coding because:
  • Ruby’s syntax is concise and “developer-happy”
  • Ruby/Rails are large, open-source codebases, improving model grounding and generation quality

What’s Previewed

• A fireside chat planned for the next day with DHH, focusing on Ruby + AI and agentic development
• Hosted by Jeremy Dar (mentioned as the chat host)

---

Segment 2: Robbie Russell — Keynote Preview on Engineering Metrics & Agent-Ready Environments

Talk Theme

Robbie Russell’s Day 1 keynote is titled along the lines of:

“Your most valuable engineer is not shipping features.”

He argues that typical metrics—such as:

• tickets closed
• bugs fixed
• PRs merged

…don’t adequately measure improvements to:

• the developer environment
• and now the agent environment (tools, workflows, long-lived knowledge, etc.)

AI Impact (As Previewed)

He claims AI changes how teams understand and manage software:

• The “heroes who know where the bodies are buried” model may hinder progression.
• With AI agents, best practices must enable agents/developers to succeed on long-lived existing applications, where institutional knowledge is often lost.

Related Resources

• Robbie’s keynote is expected to be available online after the conference (YouTube suggested).

---

Segment 3: Rails Foundation Update — AI Pillar, Agent Benchmarking, and Security

Delivered by Amanda Parino, executive director of the Rails Foundation.

AI Pillar Initiatives (Major Concrete Products)

• Agents on Rails: an ongoing benchmark to evaluate new AI agents as they are released, targeting near-real-time evaluation (goal: within ~24 hours).
  • Benchmarks include:
    • how agents write Rails code
    • time-to-complete
    • token cost
  • Purpose: help users choose which agents to use for different workflow steps, reflecting tradeoffs
• A reference library of open-source apps:
  • centralized place for agents/developers to learn from “best” Rails applications
• Communication shift:
  • “Rails is changing from prompt to IPO” (motto tied to agent workflows)

Benchmark Run Methodology Changes

• Initially ran agents at default settings
• Then expanded because users requested:
  • tests at max level
• “Max level” runs reportedly launched on Monday, with additional data added to the website afterward.

Security Direction

The Rails Foundation says it is:

• discussing a proactive, state-of-the-art security program with Rails Core
• emphasizing AI-era security as “vital,” including:
  • infrastructure funded by the Foundation
  • Rails Core retaining domain ownership for the security work itself

Education / Docs Work (Concrete, Non-Agent Related)

• Docs progress reported:
  • edited/updated all 45 guides
  • added guides including Accessibility
  • nearing merge of a Hotwire guide
  • completed multi-part tutorials (including a Cookpad case study; Robbie Russell involved)

Logistics + Community

• Mentions of Rails World app and schedule management
• Sponsors highlighted: AppSignal, Procore, Dell/Intel, Shopify, etc.
• ThoughtBot referenced for mechanical bull and live music tie-ins
• Limited-edition vinyl part of the event experience

---

Segment 4: DHH Opening Keynote — The “Age of Agents” and How Software Development Changes

This is the core keynote delivered by DHH. He uses a historical analogy (portrait painting vs photography) to argue that agentic AI represents a technology inflection point.

Inflection Point Analogy: “Brownie” → AI “Kodak Moment”

DHH compares modern AI to cameras becoming:

• mass-accessible (Kodak Brownie)
• triggering an explosion of new “form factors” and usage patterns

He also provides his “before/after” date for agent software creation:

• Nov 24, 2025: described as when agentic tooling became widely usable (for him), marking the “Age of Agents” tipping point.

The Roller-Coaster of Model Progress

• He describes cycles of:
  • excitement
  • then disillusionment (models “sucked” or regressed on quality/benchmarks)
• Later improvements reportedly restored confidence, referencing multiple model releases in subtitles, including:
  • “Fable 5”
  • “GPT-6 Astra”
  • “DeepSeek 41 Flash”

Productivity and Scale Claims

• DHH claims dramatic productivity gains:
  • In 20 months, he wrote half as much code as in the prior 21 years (as stated)
• A “lines of code” anecdote later:
  • ~30,000 production Ruby lines/year previously
  • ~150,000 lines in a month more recently
  • partly attributed to generated output being more verbose (including Rust/agent-related output)

“Pencils Down”: Writing Code by Hand Is Becoming Exceptional

• At 37signals, the team reportedly decided they are effectively done writing code by hand as a normal business practice.
• Manual coding becomes exceptional—used only when agents can’t deliver results.
• He describes early “vibe coding” by designers into Basecamp 5 as “Swiss-cheese-like” at the architecture level, followed by the corrected realization:
  • the issue wasn’t agent usage
  • it was needing to better produce integrated changes

---

Product Direction Example: Hey → Toward Native Apps + Rust Backend

Hey (Email App) Modernization Story

• Front end: shifting from web app approach to six native applications
• He claims the cost of building high-fidelity native apps drops drastically with agent-driven development

A Shopify-like parallel is cited:

• Shopify rewrote the Shop app into a native application (replacing React Native) with a small team quickly (per subtitles)

Backend: Rust for Performance/Efficiency

• He describes Rust as unpleasant to read directly, but says agents can generate it so humans don’t need to work in it manually
• Claimed resource improvements:
  • ~99% less CPU
  • ~95% less memory
  • fewer hosts needed (still using multiple for redundancy)
  • “even Raspberry Pi”-scale performance mentioned as a rough estimate

---

Workflow Pattern: Async Agent Execution (Not Chat)

DHH argues the best way to work with agents is asynchronously, like assigning tasks to coworkers:

• not sitting in chat waiting for tokens

---

Ruby / Rails “Fit” in the Agent Era

Rails is positioned as strong for agent workflows because:

• conventional patterns (“convention over configuration”)
• token efficiency and abstraction enabling “single developer goes far”

He references Rails Foundation work as well:

• updating agent evals because agents can saturate benchmarks quickly
• moving to harder benchmarks and more “elevation” in required complexity as agents improve

---

Architecture Implications

He warns traditional architecture approaches may need re-evaluation because:

• with agents acting at scale (hundreds/thousands/more processes mutating systems), bottleneck abstractions may become less effective
• synchronization and repetition costs change when agents become capable

---

Interface Expectations: CLIs and Token Economy

• He criticizes “chatbot injection”
• Encourages:
  • CLIs for AI-driven integrations
  • an emphasis that if an app lacks a CLI, it should be delivered soon (“by next Friday,” paraphrased from subtitles)

Example:

• Hey uses search via Elastic
• He claims agents can search by concepts rather than exact keywords, improving discovery (e.g., finding old emails when details aren’t remembered precisely)

---

Agent-Era OS / App Creation Demos

Omachi / Omachi-like Distribution

• He highlights Omachi (his Linux distribution; subtitles alternated between “Amachi/Omachi” inconsistently)
• Claims extremely fast install time (seconds-scale mentioned)

Agent-Generated Apps

He describes building apps with agents, including:

• calculator
• writing app
• video editor
• presentation software

Process described:

• “one screenshot of choices” → agent generates UI
• pushes to a public repo
• burns an ISO image

He frames this as shifting what’s feasible for individuals:

• small prompts can produce shippable software quickly
• resulting apps can be small (e.g., “half a megabyte” binaries) compared to older bulky paradigms

---

Cautions and Security Framing

DHH says concerns deserve debate but may be overstated. Risks he mentions include:

• security issues
• agents possibly having expiration/behavior limits
• potential misuse (“bad things”)

Countermeasures referenced include:

• “Hot Cell”
• quarantining after a Rails CVE involving an image library leak

---

Outlook: Optimism + Societal Analogy

• Uses an ATM vs bank teller analogy to argue productivity tools tend to change roles rather than eliminate them.
• Emphasizes optimism rather than doom (“p’eloom” phrasing).
• Concludes that humans can defend and guide AI use through:
  • agency
  • defensive practices

---

Main Speakers / Sources (As Identified in Subtitles)

• Andrew Noslac (Shopify staff developer; co-host for pre-show livestream)
• Adriana Chang (Shopify staff engineer; co-host for pre-show livestream)
• Yukihiro “Matz” Matsumoto (Creator of Ruby; guest interview)
• Robbie Russell (CEO of Planet Argon; Maintainable/On Rails podcast; guest interview)
• Amanda Parino (Executive Director, Rails Foundation; foundation update)
• DHH (David Heinemeier Hansson) (Opening keynote speaker)