Why let Claude have all the fun? I demoed my new library yrby-agent at @sfrubyconf, which lets an AI agent connect as a peer to your document, using @paolino’s RubyLLM behind the scenes
Starting to roll out today: Claude Whiteboards, Claude Sheets and Claude Data.
Think a plan through on a board, build a tracker in a sheet, and tell a story with your data.
RubyLLM 2.2 is out! Released it on stage at @sfrubyconf
Coding sandboxes with my new mxc gem.
Skills support with my new skill gem.
In memory, disk, and Active Storage file stores.
Tools can ask a person and wait.
Chats survive a deploy.
Tests without a provider, 23 providers.
All optimized for coding agents.
The release I'm proudest of.
https://t.co/wt3N8S6d4K
"I think Ruby's best years might be ahead of it."
Garry Tan (@garrytan) of @ycombinator opened SF Ruby Conference with words of appreciation for the community and @evilmartians.
To Garry, agents need readable, consistent code that does things the same way every time, and that's what makes Rails conventions perfect for agentic coding.
Watch the full opening:
Just fine-tuning ModernBERT after dropping DeBERTa, watching GLiNER fall over, and giving up on Julia, Kev and Laya
what are we even doing with these names
I have a rust core in a CRDT implementation which has thread safe documents and operations, and opus 5.5 wrapped every call into the documents with mutexes which serialized everything. When I pointed it out, it removed them and fixed what it found in a non blocking way
We’ll probably get there eventually, but it’s not the reality we live in right now!
I can't wait for everyone to vibe their own TLS impl. And their own OS, programming language, runtime! Definitely no bugs gonna happen. Make no mistakes.
Dependencies are so over!
People gotta realize that the whole idea of vulnerabilities being found and exposed in open source is what protects the industry from bad actors. Otherwise, they'll find ways into your app without you knowing (as they do). CVEs are a solution, not a problem. Update your dependencies!
The 🎉40th🎉 edition of Ruby AI News is live! This edition features recent security issues, Ruby’s MCP SDK reaching Tier 1 status, a closer look at building AI evals for Rails, addressing AI anxieties, and much more
https://t.co/gkyouLmG5j
The fallacy being spread about code reviews goes like this:
Guys the future is to not have code reviews. Therefore, be prepared. Therefore, don't do them NOW.
No, now is not the future.
Hey friends, send us some work! We're smart, cool, and our job is to make you succeed 😎
Come work with @evilmartians Martians on all these things and more:
Rails performance and CI optimization
LLM benchmarks and agent infra
Product and design for devtools and builder tools
Product and design for cybersecurity
Forward-deployed engineering for AI-native products
UI coding harness: make agents stick to your design system
Ruby, Go, TS SDKs and adapters
Secure, agent-ready npm packages (provenance)
Cloud and DB migrations
Inspired by @fractaledmind’s berlin.rb talk, I wrote judgelint, a tool that lets you express nuanced project rules, and have a system one model (like Jev) judge whether your code complies with it. The goal was something fast and cheap enough to run as part of CI, and integrate nicely into GitHub.
Give it a try! https://t.co/cUEIvkYjwW
Jev is the default, but it supports lots of others (like clef and the OpenAI decisions api). Doesn’t support local models atm, but happy to take a PR for it!
🎟️ I'm giving away free tix! Tmrw's the last day. Tech/CS students in the Bay Area, come join me at the San Francisco Ruby Conference, Nov 10-12. Learn about the startups and technologies powered by the Ruby programming language.
@samokhvalov This is pretty awesome, and i've been digging into it from the Ruby side. Are you open to ruby PRs right now on the work you've done so far?