My experiments with agent-augmented dev feel like "coding by dictation", and always require manual review and MANY follow-ups.
Code style: terrible.
Naming things: terrible.
Math: terrible.
Forward-thinking design: terrible.
Unsure if I'm more productive or less. 🫤
JRuby guy, Java, Ruby, OSS business liaison.
Find me: @[email protected] or @headius.bsky.social
Sponsor me: github.com/sponsors/headi…
- I occasionally want to close an experimental PR without losing the context and changes associated with it. Is there a better way than dumping a patch into comments as seen below? I don't want lingering topic branches.
- Is there any standard in place for marking content with its level of AI contribution? Basically, a *standard* way to mark your content as fully human generated all the way up to fully AI generated. Doesn't seem like there's a good standard.
- There are few conferences I have been to with such engaged and excited developers. Definitely worth your sponsorship.Shape the future of Ruby! Become a sponsor at #RubyConfIndia2026 and align your brand with pioneering minds, passionate Rubyists, and transformative tech in the Ruby ecosystem. Discover more: rubyconf.in/prospect/RCI_P… #RubyConfIndia #RubyOnRails #RubyConf #RubyConf2026
- I'm about ready to give up on GitHub Milestones altogether. * No linkage with releases or tags whatsoever * No way to mark multiple milestones * Crappy management UI What am I missing?


