I build, ship, scale or AI-ify Rails & web apps. 21y engineering, learning by contributing to OSS. Trust AI, verify always.
Remote · Worldwide
Joined January 2008
- Tweets4.9K
- Following338
- Followers424
- Likes594
Pinned Tweet
I've loved Ruby on Rails since early 2008, and after all these years, my very first PR to Rails itself was merged just last week. Such a great feeling!
Lately, I've also been contributing to my favourite OSS RubyLLM and ActiveAgent a lot.
I believe one of the best ways to learn in the AI era is to get deeply involved in open source. There's so much to learn from building, contributing, and collaborating with great engineers.
Always happy to connect with fellow Ruby/Rails enthusiasts. 🤝 #ruby #rails #oss
This week there were 46 contributors to Rails.
Thank you to all those whose PRs were highlighted this week:
@thomasklemm, Kengo Johira (kehra), Islam Elsayed (IslamElsayed), @kamipo, Wei-Lun Li (b84330808), @kuokkanensampo, Clément Joubert (czj), Richard Desmarais (fuzzyrichie), Masaki Ozawa (ozamasa), Jeff Tsang (JeffTsang), @linyiru, Jason Max Lee (jasonmaxlee), @cestercianAf, @adriannakchang, Mick Zijdel (mickzijdel), Jonathan del Strother (jdelStrother), Greg Pavlik (g-pavlik), Wleeaf (wleeaf), @svyatov, Andrew K. Fader (andrewfader), Andrew Quinn (anquinn), @nganpham, Carlos Daniel Pohlod (carlosdanielpohlod), Simon Isler (sislr), Martin Samson (masom), & Sean Doyle (seanpdoyle).
Check out the full list of the week's contributors at: contributors.rubyonrails.org…
1
3
187
New model, new starting point.
Every time we have a new model, I run the same simple prompt on all my projects:
"Read this entire project and find everything that’s broken, inconsistent, inefficient, or no longer needed. Fix what you’re sure about and list the rest for me to review."
highly recommend trying
28
Like coding, AI makes writing cheap as well. Original experience, strong evidence, and human judgment are more valuable than ever.
My new blogging experiment: 👇
• Brief mode for quick reads
• Details mode for deep dives
• Sources & tests for proof
• AI integration for interactive learning
Blogs shouldn't just be read. They should be explored.
Learning MCP by fixing it: nine RubyLLM pull requests, and what each bug taught me about MCP. By @yorzi wangyaodi.com/blog/learning-…
10
Learning MCP by fixing it: nine RubyLLM pull requests, and what each bug taught me about MCP. By @yorzi wangyaodi.com/blog/learning-…
Don't worry, let others worry!
If you want to get out of your current life situation, you shouldn’t be worried about how long it will take to change it. You have nothing better to do than to work on something meaningful. Who cares if it takes 12 months or 12 years when it's going to be difficult either way.
1
61
I just figured out why there were some replies which I was not notified. Door is open now, welcome to discuss anything about Ruby/Rails/AI/OSS/SAAS/tools. 🤝
1
1
68
One practice that's worked well for me: I ask AI to maintain two kinds of docs.
For agents: precise specs, constraints, and execution rules.
For humans: intent, decisions, and what changed.
Different audiences, same source of truth.
Both should evolve with the code and tests, so humans can review what matters and stay accountable.
3
3
96
Related research on how stale instructions affect coding agents:
arxiv.org/abs/2610.30571
A useful reminder that documentation quality matters as much as instruction quality.
1
52
Human attention is expensive.
We shouldn't have to reconstruct every step an AI agent took.
What matters is understanding why it made certain decisions, what changed, what risks remain, and where human judgment is needed.
Good documentation shouldn't just explain what AI did. It should help humans make better decisions with less effort.
AI handles execution. Humans own the outcomes.
17
"could this be simpler?"
Same question Rails taught me in 2008. Agents just made it cheaper to ask.
Left alone will introduce another layer, another flag, another file. The diff grows. The review slows.
Asking costs one line. Not asking costs the reviewer. teapaw.org/dhh/status/21187129242…
The only repeated redirecting prompt I typically type or dictate from here is "could this be simpler?". Yields improvements every time. Seems like it should be part of the model system prompt by now.
45
People follow for many reasons, but unfollow when they no longer see enough value (for him/her).
They're called followers, not friends, for a reason. Unfollowing shouldn't feel personal.
I'd love X to recommend accounts to unfollow, not just follow, with reasons why. @elonmusk
A better social network shouldn't just help us follow more people. It should help us follow the right people.
Attention is limited. Let's spend it wisely.
teapaw.org/wtffstories/status/211…
Quick question, @X family
Why do people follow and then unfollow?
Content not interesting, too many posts, or no follow-back? 😂
Be honest. What's your reason?
2
4
193
Smaller accounts aren't automatically better. And follower count doesn't define how interesting someone is.
Let's connect if you're:
- building SaaS
- contributing to OSS
- working with Ruby on Rails
- using AI daily
- building something on your own
- sharing niche thoughts and findings
I'd love to hear what you're working on. @X
2
22
Andy Wang retweeted
Not money, not time - but attention.
“The real currency of life is attention
What you choose to pay attention to
And what you do about it”
- @naval
THE REAL CURRENCY OF LIFE
OUT NOW